Module – 5: Introduction to Media Research

5.1 Application of Media Research

5.1.1 Research in Print Media: Circulation Research , Readership Research ,
Readability Research
5.1.2 Research in Electronic Media : Rating Method, Non- Rating Method
5.1.3 Research in Advertising : Pre testing Methods: Direct Ratings, Portfolio
Tests
5.2 Testing Methods
5.2.1 Physiological Tests: The Eye Camera Test
5.2.2 Physiological Tests: The Pupillo-meter Test
5.2.3 Physiological Tests: The Tachisto Scope Test.
5.2.4 Consumer Jury Test
5.2.5 GSR Test
5.2.6 Post Testing Methods: Enquiry Test
5.2.7 Post Testing Methods: Recall Test
5.2.8 Post Testing Methods: Recognition Test
5.2.9 Post Testing Methods: DAR Testing
5.3 Statistical Application in Media Research
5.3.1 Statistical Application in Media Research ; Statistics Applied in Research,
Frequencies and Percentages,
5.3.2 Measures of Central Tendency: Mean, Median and Mode;
5.3.3 Measures of Dispersion: Range, Standard Deviation and Mean Deviation;
5.3.4 Association & Simple Correlation.
Unit – 5.1: Application of media research

At the end of this unit, you will be able to understand:

 Research in Print Media: Circulation Research , Readership Research ,  Readability Research

 Research in Electronic Media : Rating Method, Non- Rating Method
 Research in Advertising : Pre testing Methods: Direct Ratings, Portfolio Tests
Introduction

Media research is the learning of social, psychological and physical aspects of the different mass media.
For instance, how much time do public spend with a particular means? Whether it has the effect of
bringing about changes in the perspectives of public? Does the use of medium have any damaging
effects? What is it that media users wants and expect to hear or read or see and experience? In this
connection it is also researched whether a medium can provide information and entertainment to more
and different types of community. Media research includes study about the development of media,
achievement and effects. It includes the methods used in collecting and analyzing information about
newspapers, magazines, radio, TV, Cinema or other mass media. It also concerns with and long-drawn-
out discussion of the scientific methods of research. It also helps to recognize the ways in which media
can meet the needs of the audience. Whether it can provide information and amusement to more and
different types of people. New technological improvements which helps to improve or enhance the
medium. Hence in order to deal with social and political issues, management and regulation of media is
needed. Unbiased evaluation of data can be achieved using media research.

There are a lot of media related research which are progressed for practical application purposes. From
the 50’s and 60’s, advertisers have been using media research to work out ways to convince potential
customers to buy products and services. Consequently, a large number of media studies were conducted
on communication efficiency. Demography and volume of audience, role of advertising in achieving
higher degree of acceptance and sell, frequencies of message to convince potential customers and
selection of media that best matched to reach the target audience were some of the advertising related
media research areas.
Media research can help us be familiar with the ways in which media can meet the needs of the
audience. Answer to the above query can help us manage media organizations much better from a
business point of view. This can make media more socially responsible, mainly if communication
professionals and media consumers become more aware about the problems the society is facing at
present. It will help them make better decisions not only to manage and regulate the media effectively,
but also to insightfully deal with social and political issues. It reflects and helps shape our perception of
the appropriate roles of media professionals, media institutions, groups and community. It helps teach
us what issues are major. It provides us with satisfactory information to form opinions. The influence of
media on modern society is very powerful but it is not in one direction.

5.1.1 Research in Print Media: Circulation Research , Readership Research ,  Readability Research

Print media is one of the oldest and fundamental forms of mass communication. It includes newspaper,
weeklies, magazine, monthlies and other forms of printed journals. Print media forms an important part
of media industry which is getting competition by electronic media and new technology called e-
newspaper applications. But still, newspaper holds its value even today and competing powerfully with
technology backed media. Basically said, words printed with ink on paper that is still attached in
different ways provide different kinds of Information, Education, and Entertainment. These may consist
of chapters or themes, pictures, advertising and so on and the emphasis is made on make-up and an
attractive appearance. Newspaper is the most accepted media in the print category. Today reading
newspapers is routine activity for the educated people all over the world. Topics/article displayed in the
newspapers are normally talked about and discussed. Print media utilize the printing means to spread
some news, information or message among a large number of people. Hence, this paper aims to
highlights the effectiveness of Print media newspapers among the consumers in Vellore city.
Printing is a process for reproducing texts and pictures, typically with ink on paper using a printing press.
It’s often carried out as a large scale industrial process and is essential part of publishing and transaction
printing. Print media make use of printing means to spread some news, information or message among
a large number of people. Print provides a broader range of access to the knowledge about the
particular topic and enables future generations to build up directly on the intellectual achievements of
earlier ones without the changes arising from verbal traditions. Print media today is challenged by its
electronic counterpart but there are several unique advantages that help both the mediums to create a
niche place in media Industry.

Print media Research may cover many things, like start with Readership, i.e. what is its quantum
Readership, Who reads it, what are the places it has readership in and so on. Other research can be on
the type of its topics and editing, like, what ratio of specific topics it covers. The articles are creative,
topic based or plain news and so on. Advertisement ratio can also be studies, for example, what are the
types of Advertisements it has? What is the ad quantum? And so on. Print media Research will all
depend on the purpose that the research work is to be done.

Let’s see the researches involved in print media:

Circulation Research: The term circulation can be explained as the number of copies of a magazine or
newspaper sod through paid subscription and other sales by mail or single copies. Circulation research
largely focuses on the demand of print medium in different markets and its delivery and pricing system.
There are two types of circulation research used in print media. The first kind considers a particular
group of readers as a group of analysis and in the second type, the unit of analysis will be the individual
reader.

Circulation has forever been a critical part of the business of journalism as well as the practical reason
for producing journalism in the first place. There are many examples of this throughout the history of
journalism as well as today.

When the first mass media outlets in the form of printers who owned printing presses raised the one
thing that kept them down was lack of circulation. Due to a lack of literacy and the expense of printing
circulation was low so there were not many newspapers. As printing became cheaper and printed
materials were more accessible to the common man, literacy rose. With the rise in literacy came a
hunger for printed information, giving the printers the circulation they needed to run weekly or even
daily newspapers.

Today we have seen the same factors that originally led to an increase in circulation now cause the
opposite. Technological advances like the Internet and social media have actually lowered circulation
among newspapers. As the news business has become more and more competitive, online editions and
the free availability of news has increased. This increase has caused users to turn away from traditional
sources for the cheaper alternative.

Primary circulation: The first type of circulation research uses a specific group of readers as its unit of
analysis. It tries to measure circulation in terms of overall characteristics of a particular market.
Circulation research can also be used to identify other market level of market structure that has an
impact on circulation market size and location showed a stronger relationship with circulation.

Secondary circulation: This type of circulation research uses individual person who reads as the unit of
analysis to determine the effects of certain aspects of delivery and pricing system on reader’s behavior.
Magazine publishers frequently conduct this type of circulation by drawing samples of subscribers in
different states and checking on delivery dates of their publication and its physical condition when
received.

Readership Research:
This kind of research started in the United States immediately after World War II. They conducted a
personal interview in which the respondents are asked to recognize the article they had read from a
selected newspaper. But a comprehensive study of newspaper readership was undertaken by American
Newspaper Publishers Association. The readership research is important in newspaper industry as it
helps to get the thorough information about the readers and accordingly they could change the content
of the publication, if required. Moreover, the rising competition from traditional and online media
makes the readership research more important.
Readership research is another method to inspect print media practice, focusing normally on magazines
and newspapers that appear every so often. Of primary importance in this context are readership
analyses that set up findings on print-media coverage and readership structure. These methods are
supplemented by reception analyses, which investigate reading habits in general, sense.

A difference can be drawn here among readership research as media advertising research, which deals
with the performance of a print medium as an advertising medium, and as editorial readership research,
which is meant at optimizing newspaper and magazine content and/or layout. The lion’s share of
applied readership research focuses on optimizing media setting up for advertising purposes as well as
optimizing the content and layout of print products. In comparison, academic reception research, which
is designed to determine more fundamental insights into readership, plays a relatively minor role. The
dominance of media advertising research in practice can be explained by its far-reaching economic
significance.

Newspaper or magazine circulation says little in itself about the number of people a print medium
actually reaches. The reality of the situation is far more complex: many copies are printed but
distributed free of charge rather than sold or are not distributed at all, other copies are sold but are not
read and, above all, a great number of copies are used by more than just one person, e.g., at home, at
workplace, or in waiting room.

What a “Reader” is?

Since circulation data fails to cover the complete picture, it is necessary to conduct readership surveys
based on broad samples. In order to do this, it is essential to define exactly who qualifies as a reader.
Should only the people who read a publication from cover to cover very carefully be counted or do
people who just flip through a publication, stopping only to read a few headlines or take a quick look at
a few pictures also qualify as readers? What if a publication is read more than once: do all reading
actions count? Does it only count if a publication is read on one specific day or does it count if it was
read at any time within the issue period or even well beyond the issue period, for example, at some time
in the last three months? Clear definitions of who counts as a reader are essential as they provide the
advertisers as well as the publishing houses of advertising space with what equates to a mutually
acceptable currency.

The most common of such currency is Average Issue Readership (AIR), for example the number of
people who have read or looked at some or all of the average issue of a publication. Just flicking through
is enough: it is not necessary to have read a publication through carefully to have come into contact
with an advertisement placed there. 

Average Readers per Copy (ARPC) can be calculated by dividing average-issue readership by average
circulation. There are many different ways of measuring average-issue readership. The four main
techniques namely: “Through-the-Book,” “Recent Reading,” “First Reading Yesterday,” and “Readership
Diaries.”

Because large national readership surveys often includes quite a lot of hundred different newspaper and
magazine titles, it is common practice to pose screening questions at the start of interviews to reduce
the title load per respondent by excluding those titles respondents “only know by name,” or titles that
are completely “unknown” to them. Subsequent questions, usually beginning with the frequency
question followed by the frequency question, are only posed on the remaining publications. Logo cards,
whichever in black and white or in color, are the main form of visual aid employed in face-to-face
readership interviews to assist respondents’ recall, as well as to help avoid title confusion. Print titles
and categories (dailies, weeklies, fortnightlies, monthlies, etc.) are often rotated to counterbalance
possible order effects. In many readership surveys, respondents are then posed a number of
supplementary questions to establish, for example, how long, where, and on how many days reading
took place, or to ascertain their “relationship” with the publication read, e.g., by asking about the
amount read (“all/nearly all,” “over half,” “about half,” “less than half,” etc.). In order to establish
reader “involvement,” respondents are also often asked to gauge how much they would “miss” a
publication if it were no longer published.
Basic contact data (having “read or flicked through” a newspaper or magazine) shows advertisers little
about the potential effectiveness of their adverts. In contrast, data on the quality of reading, allows
conclusions to be drawn about the chances readers have of coming into contact with advertisements,
which is a prerequisite for advertising effectiveness. Respondents who have read a publication carefully
from cover to cover, for example, are more likely to come into contact with a particular advertisement
than those who have only taken a look inside, flicked through briefly.

In summing up his comprehensive review of methods developed in the field of readership research,
Michael Brown states there is no “gold standard,” no single, inalienable currency that provides an
equally valid measure of readership across the entire spectrum of print titles and categories:

“All methods have differing advantages and limitations. Arguments as to methods’ ‘validity,’ in any
absolute sense, are arid; they should be judged by their ability to deliver readership estimates which
allow comparisons between different newspapers and magazines which are minimally biased” ( Michael
Brown 1999).

Provided that there are no universally accepted methods, it will continue to prove difficult to further go
with many different techniques being employed to estimate readership around the world today, as well
as to establish common methodological standards. This is important, however, for international
advertising planning in an increasingly globalized world economy.

To know who reads the media in terms of their profile and likeness of the media, especially there are six
types of studies:

a. Reader profile - Reader profile studies provides a demographic summary of the readers of a
publication. At present, psychographic studies and lifestyle segmentation research are
conducted to more about readers. These data can be used to focus the content of the
publication, arrange advertising campaign and increase subscriptions. One cannot predict the
nature and degree of newspaper reading among individuals with this demographic data. In
psychographic studies, the reader has to point out whether they have the same opinion or
disagree with a group of attitudinal statements.

Afterwards, patterns of response are analyzed to observe how they connect or cluster together.
People who show good levels of agreement with questions that cluster together can be
described as labels that sum up the substance of the question. The psychographic and lifestyle
segmentation studies provides an added insight about editorial aims, target audience and
circulation goals.

b. Item-selection studies - This study is used to determine who reads specific parts of the
newspaper. In item selection studies, aided recall, measurement used to find out the readership
of a particular thing is used. Here, the researcher shows a copy of newspaper to the reader to
find out which are the stories they remember.

Instead of conducting personal interviews, the researcher can also use telephonic interview to
collect data. In an item selection study, the unit of analysis will be a precise news article, or a
specific content category. The readership of these categories is then related to certain audience
demographic or psychographic characteristics. For example, a research study reported that
juvenile readers have reading habits different from adults.

c. Reader nonreader studies - This type of study looks to find out the reasons people do not
read the newspapers. It can be executed by personal, telephone or mail interviews with minor
modifications. It is hard to define the term non-reader. In some studies, a non reader is
determined by a no answer to the question, “do you generally read a newspaper?” Otherwise
more specific questions can be asked like, “have you read newspaper today or yesterday?”. A
third form of this question includes using multiple response categories.

d. Uses and gratification studies - To study all media content. For newspaper it is used to
determine the motives that lead to newspaper reading and the personal and psychological
rewards that result from it.
 This type of readership research is widely used in study of all media content. The respondents
are provided with a list of possible uses and gratification and are asked whether any of these are
the motives behind their reading. Then, the researcher concludes the responses and an average
score for each motivation item is calculated. The main aspire of this study is to find out the
motives that lead to newspaper reading and the personal and psychological rewards that result
from it. Gratification from reading the newspaper seemed to differ across cultural group.

e. Editor-reader comparisons- Here, we are inquiring a group of editors about some topics and
these answers are compared with those of their readers to find out whether there is any kind of
resemblance between their opinions.

This kind of study reveals whether the space allotment of the newspaper made by the editor
matches the public interest. An editor reader comparison study revealed that both groups
agreed on the importance of many journalistic standards, but readers did not value professional
staffing goals and enterprise reporting as highly as the editor did.

Readability Research: Readability is a word used to determine the ease with which public read and
understand a particular wording. It is one of the most important factors that represent the directness of
the concerned text.
“Readability is what makes some texts easier to read and understand than others”.
The idea is different from “legibility”, which is concerned with typeface and layout. According to George
Klare (1963) readability is

“the ease of understanding or comprehension due to the style of writing.”

This explanation focuses exclusively on writing style which focuses on the semantic and syntactic
elements and excludes such issues as content, consistency, and organization within the written text.

As per Edgar Dale and Jeanne Chall (1949): “The sum total (including all the interactions) of all those
elements within a given piece of printed material that affect the success a group of readers have with it.
The success is the extent to which they understand it, read it at an optimal speed, and find it
interesting.”
The source of the initial readability study was the concept of studying literature from a statistical view
point by English literature professor, L.A. Sherman in the 1880s. He discovered that using shorter
sentences and concrete requisites increases the overall readability of the text.
Psychologist Edward L. Thorndike of Columbia University published The Teacher’s Word Book in 1921
which programmed 10,000 words by frequency of use. Thorndike found that the more regularly a word
is used, the more familiar it is and the easier to use. This concept of vocabulary frequency lists proved to
be a great help for teachers to evaluate the readability of reading materials for their students.
Increasingly in USA, the adult literacy studies were the first studies conducted to evaluate the reading
ability of the general readers.

The respondents were rated accordingly on the basis of their different reading abilities. Still, the main
worry of educators, writers, journalists, corporations and government agencies was the issue of a lack of
a generalized method for measuring readability of a particular text. Afterward, a series of research
studies were conducted by a community of readability scholars such as Edgar Dale and Ralph Tyler
(1934), Bernice Leary and William S. Gray (1935), Irving Lorge (1938), Rudolf Flesch (1946, 1949, 1964,
1979), Edgar Dale and Jeanne Chall (1948), Robert Gunning (1952), Wilson Taylor (1953), George Klare
(1963, 1975, 1976, 1980), G. Harry McLaughlin (1968), Edward Fry (1963, 1968, 1969, 1977) and lots of
more. These scholars were credited for increasing formulas for measuring the readability of a given text
and it proved to be a boon for those groups of people who were really concerned about the readability
factors in their texts. Some of the normally used readability formulas are Flesch Reading Ease Score,
Flesch-Kincaid Grade Level, SMOG Readability Formula, Fog Index and Fry Readability Graph. Full-grown
research in readability began in the 1920s. There was a great deal of development in research in the
area of readability from 1920 to the middle of the 1990s. According to Gilliland (1975), matching the
reader and the text has come to be called “readability”. In the past, research on readability has focused
on the growth of practical methods to match reading materials to the abilities of students and adult
readers.

For the start of the year of the study of readability, researchers believed that reading difficulty was
related to reading material. This research focused on devising procedures and instruments that would
reliably and validly distinguish easier from more difficult reading materials. Therefore, the study of
readability was concerned with the search for factors in reading material which could be easily counted
and measured. The outcome of these investigations led to the formulation of “readability formulas”
which were used to count measure and quantify the difficulty level of a given text. The quantity of
readability by using readability formulas were quantitative approach.

They focus exclusively on linguistic factors such as words and sentences, like, number of words and
sentences, word complexity, word and sentence length which can be easily measured and quantified. As
such, this shows that readability was completely conceptualized from a paradigm found outside the
reader where reading difficulty or comprehension were predicted by looking through reading material
instead of looking through the reader. But as researches in readability gained foundation, researchers
counter depicted that linguistic structure cannot be the only criteria for knowing the readability of a
particular text. They were of the view that involving the readers of the text is also one important feature
for measuring the complexity level and readability of the text. As the idea of measuring readability is
meant exclusively for the readers, it is very important to know the reader’s varying reading skills for
depicting how much readable a concerned text is.

Readability research became one of the most important and valuable considerations in mass
communication process, like in the printed sector. Various forms of print media such as newspapers,
magazines, journals and books require predictions of readability of their printed text prior to publication
in order to make their audiences understand the printed materials easily. If a printed text is not readable
sufficient, the purpose of writing in its first place will be defeated. The writing needs to be linguistically
planned in such a way that it fits the reading skills of the intended audience. Hence, the issue of
readability factor in printed media is a grave concern for journalists, writers, editors, publishers, and all
those who are involved in the print media industry in order to increase the readership and circulation of
their printed materials. Many researchers have also been reported where increasing the readability of
the printed materials directly leads to a sheer increase in the readership and circulation of their printed
medium.
There are wider and narrower senses in which the term readability can be taken. In the narrower sense,
it refers to the development and use of readability formulas and related objective methods which use a
small number of measures of variables such as average number of words, syllables, etc., in a sentence or
text. This is where readability research has started. In the wider sense, readability is discussed by
inculcating the qualitative approaches to its study. This involves the addition of readers’ analysis in the
determination of readability of a particular text. Through a series of reviews on readability researches
conducted on a wider sense, the paper will explore interesting insights into different aspects of
readability researches. Moreover, the paper will also feature a critical evaluation of readability by
analyzing the text-specific objective (quantitative) and reader-specific subjective (qualitative) testing of a
literary text. This will take into consideration the efficacy of the readability formulas in regard to reading
grade and subject specific reader analysis.

Research in readability has gone through tremendous changes during the last 50 years. Quantitative
measurement of written text by using readability formula is not enough to provide a holistic evaluation
of readability (Oliver et al., 1998). A qualitative technique of readability assessment by analyzing certain
characteristics of the reader such as reading skills, knowledge and interest level, purpose of reading, etc.
is found to be acceptable by many researchers. The new transformation in readability research tends to
focus on what is happening in the readers’ mind during reading (Janam et 54 Bidyarani Asem al., 2010).
Readability, with a holistic consideration should always be the “ease of reading” created by the choice of
content, style, design, and organization that fit the prior knowledge, reading skill, interest, and
motivation of the audience.
5.1.2 Research in Electronic Media : Rating Method, Non- Rating Method
Research in the electronic media has been a fastener management tool for successful managers for
several decades because managers need to know what their target audience wants.
There are a lot of types of researchers in the electronic media who provide the quality information that
is needed. Take an example of, there are professional research companies; in-house research
departments in most radio, television, cable, TV network, and satellite operations; and many research
departments or divisions in colleges and universities. Electronic media research is a million dollar
business that frequently changes because of advancements in technology and improved research
methods. This section introduces some of the more widely used research procedures in this area.

Broadcast research has developed speedily in complexity and volume since its beginnings in the 1920’s.
In the early years of broadcasting, the broadcasters were experimenters and hobbyists who were
attracted mainly in making sure that their signal was being sent and received. The popularity of radio
was unheard of, and there was no reason to be concerned with audience size at that time.

The circumstances changed quickly during the 1930s as radio became a well-liked mass medium. When
radio began to attract big audiences, concern emerged over how radio would be financed. After much
discussion, advertising emerged as the choice over government financing or taxes on sales of
equipment. The recognition by radio listeners of advertising on radio was the first step in the
development of electronic media research.

Advertisers were the initiators of broadcast research. When commercials began airing on radio stations,
advertisers obviously wondered how many listeners were exposed to their messages and how effective
the messages were. It became the accountability of broadcasters to provide empirical information about
the size and characteristics of their audience. This situation still exists—advertisers constantly want
information about the people who hear and see their commercial announcements.

In addition to information about audience size, advertisers became interested in why public behave the
way they do. This led to the growth of the research area known as psychographics. Though, because
psychographics data are rather vague, they were not adequate predictors of audience behavior and
advertisers wanted more information. Research procedures were then designed to study lifestyle
patterns and how they affect media use and buying behavior. Such information is important in designing
advertising campaigns: If advertisers know the lifestyle patterns of the people who purchase their
products, they can design commercials to match those lifestyles.

Electronic media research studies fall into two broad categories: ratings and nonratings research.

Rating Research:
When radio became popular and advertisers began to grab its potential for attracting consumers, they
were faced with the problem of documenting audience size. The print media were able to provide
circulation figures, but broadcasters had no equivalent information, just unproven estimates. Early
attempts at viewer’s measurement failed to provide adequate data. Telephone calls and mail and to the
radio station from listeners was the first sources of data, but it is a well-known fact that telephone calls
and mail from a handful of listeners or viewers do not represent the general audience. A fable that
someone started long ago, that many people still believe today, is that one letter or one telephone call
to a radio station is equal about 10 listeners. Advertisers and broadcasters promptly realized that more
information was urgently needed.
Since 1930, when a group was called the “Cooperative Analysis of Broadcasting” conducted one of the
first audience surveys for radio, several companies have attempted to provide syndicated audience
information, but two companies have survived and provide the bulk of syndicated ratings information
for radio, television, and cable: Nielsen Media Research for television and Arbitron for radio. The United
States is divided into about 363 metropolitan areas. Nielsen conducts TV ratings in 210 of the markets;
Arbitron conducts radio ratings in 287. In many markets, both companies provide ratings data all
through the year, known as continuous measurement, not just during certain times of the year as in the
past.
Ratings Methodology
The research methodologies which are used by Arbitron and Nielsen are complex, and both company
publishes several documents describing its methods and procedures that should be consulted for
specific information. The data for ratings surveys are at present gathered by two methods: diaries and
electronic meters/recorders. Each method has its advantages and disadvantages.

Each ratings service uses a formula to set up a minimum sample size required for a specific level of
statistical efficiency, but there is no assurance that this number of subjects will actually be produced.
Though many people may agree to participate in an viewers survey, there is no way to force them all to
complete the diaries they are given or to use electronic meters accurately. In addition, completed
diaries are often rejected because they are illegible or obviously inaccurate. The companies are lucky to
get a 50% response rate in their local market measurements.

Data Collection:
1. Possibly the best-known method of collecting ratings data from a sample is by means of
electronic ratings gathering instruments, in particular the Nielsen People Meter, as mentioned
earlier in the section. The People Meter was earlier introduced as the audimeter in 1936 to
record radio use on a moving roll of paper. In 1973 the audimeter altered to a new technology
and the device was called the storage instantaneous audimeter (SIA). At last, during the 1987–
1988 television seasons, Nielsen offered service with the new People Meter. The audimeter
records viewing data and the central computer in Dunedin, Florida, calls each NTI household
each day to retrieve the stored data. These data are used to calculate the National Television
Index. The data collection is automatic and does not require involvement by anyone in the NTI
households.

2. The 2nd major form of data collection, requires subjects to record the channels they tune to or
the stations they listen to, when they watched or listened such as “prime time” (8:00 p.m.–
11:00 p.m. EST). Arbitron uses diaries for radio; Nielsen uses diaries for the households in its
NAC (National Audience Composition) sample to increase the information gathered from the
metered households because the audimeter cannot record the number of people who are
watching each television set.

3. The 3rd foremost technique used to collect data, the telephone is not currently used by any
accredited ratings research company. However, Arbitron and Nielsen do use the telephone for
initial sample recruiting and to conduct a variety of special studies, allowing clients to request
 roughly any type of survey research project. One of the most regular types of custom work is the
telephone coincidental. This procedure measures the size of the medium’s spectators at a given
time; the survey coincides with actual viewing or listening. This method involves selecting a
sample of households at random and calling these homes during the viewing or listening period
of interest. Respondents are just asked what they are watching or listening to at that moment.
This method avoids having respondents try to recall information from the earlier day.
Coincidentals are inexpensive (generally a few thousand dollars) and are often used by station
management to receive immediate feedback about the success of special programming. In most
cases, coincidentals are used for advertising sales purposes.

4. The 4th method of ratings data collection is Arbitron’s Portable People Meter (PPM), which the
company introduced in the mid-1990s to advance the validity and reliability of radio ratings.
Arbitron describes the PPM in its publication, PPM Encoding Handbook for Radio, in this way:
 The PPM measures all kinds of radio listening, including listening to radio stations on the Internet.
Though, for the PPM to recognize a radio station, the station’s signal must be encoded. As per
Arbitron “A radio station (FM/AM/HD Radio/ Internet) whose broadcasts include unique inaudible
codes that can be heard by the meters carried by PPM panelists. Only encoded stations can be
measured in the PPM service. A station is not required to subscribe to the PPM data in order to
encode.”

Since 2014, Arbitron’s PPM is used in quite a few markets in the United States, but the methodology
continues to be the focus of heated debate among broadcasters, advertisers, and even government
officials.

Interpreting the Ratings

Interpreting broadcast ratings and understanding the terminology used in can best be explained
with an example. As this example uses television, the procedures are the same for radio ratings.
Let’s suppose that Nielsen has collected TV viewing information (see below Table 1) for a specific
day part on “traditional” network television.
Remember that Nielsen’s NTI sample includes about 10,000 households in the United States, and
the data collected from them are generalized to the total population of about 114.5 million
television households. The first number to calculate is the rating for each network.

Table 1: Hypothetical TV Viewing Data

Networks HouseHold
STAR 2200
SONY 2000
NGC 1800
Now Watching 4000
Total 10000

Rating. An viewers rating is the percentage of people or households in a population with a television or
radio tuned to a specific station, channel, or network. Therefore, the rating is expressed as the station or
network’s audience divided by the total number of television households or people in the target
population:
Above formula use is typical in mass media research. Nevertheless, the numerator in the formula is
actually People or Households Viewing TV / People Listening to Radio. For example, using the
hypothetical data in Table 1, STAR’s rating is calculated as:

This indicates that around 22% of the sample of 10,000 households was tuned to STAR at the time of the
survey.
The mutual ratings of all the networks or stations during a specific time period (daypart) provide an
estimate of the total number of Homes Using Television (HUT). Since radio ratings deal with public
rather than households, the term Persons Using Radio (PUR) is used. The HUT or PUR can be found
either by adding together the households or personnel using radio or television or by computing the
total rating and multiplying it by the sample (or population when generalized). The total rating in the
sample data in Table 1 is .60, which is calculated as follows:
In other words, about 60% of all households (HH) with television were tuned to one of the three
networks at the time of the survey. As mentioned, the HUT can also be computed by multiplying the
total rating by the sample size: 0.60 3 10,000 5 6,000. The same formula is used to project to the
population. The population HUT is computed as: 0.60 * 125.5 million 75.3 Million.

TV stations, networks, and advertisers naturally wish to know the estimated number of households in
the HUT tuned to specific channels. Ratings data from the sample of 10,000 households are used to
provide a rough estimate of the households tuned to each network (or station).

Share. An audience share is the percentage of the HUT or PUR that is tuned to a specific station,
channel, or network. It is determined by dividing the number of households or persons tuned to a
station or network by the number of households or persons using their sets:

From Table 1, the sample HUT is 6,000 (2,200 + 2,000 +1800), or 60% of 10,000. The audience share for
STAR would be:
That is, of the households in the sample whose television sets were on at the time of the survey, 36.7%
were tuned to STAR. (People may not have been watching the set but may have recorded that they did
in a diary, or the People Meter recorded the information.) The shares for SONY and NGC are computed
in the same manner.

Related Ratings Concepts:

Ratings, shares, and other figures are computed for a variety of survey areas and are split into several
demographic categories. For an additional fee, ratings companies also provide custom information such
as ratings in specific zip codes. Although ratings and shares are important in audience research, a
number of other computations can be performed with the data.

A metro survey area (MSA) corresponds to the Consolidated Metropolitan Statistical Areas (CMSA) for
the country as defined by the U.S. Office of Management and Budget. The MSA generally includes the
town, the county, or some other designated area closest to the station’s transmitter. The designated
market area (DMA), another area for which ratings data are gathered, defines each television or radio
market in exclusive terms. (At one time Arbitron used the term Area of Dominant Influence, or ADI, to
describe the DMA but has since changed to Nielsen’s designation.) Each county in the United States
belongs to one and only one DMA, and rankings are determined by the number of television households
in the DMA. Radio ratings use the DMAs established from television households; they are not computed
separately.

The total survey area (TSA) includes the DMA and MSA as well as some other areas the market’s
stations reach (known as adjacent DMAs). Broadcasters are most interested in TSA data because they
represent the largest number of households or people. In reality, however, advertising agencies look at
DMA figures when purchasing commercial time for television stations and MSA figures when purchasing
radio time. The TSA is used infrequently in the sale or purchase of advertising time; it serves primarily to
determine the reach of the station, or the total number of people or households that listened to or
watched a station or a channel. Nielsen’s term NSI area is equivalent to Arbitron’s TSA. Ratings reports
contain information about the TSA/NSI, DMA, and MSA. Each area is important to stations and
advertisers for various reasons, depending on the type of product or service being advertised and the
goals of the advertising campaign. For instance, a new business that places a large number of spots on
several local stations may be interested in reaching as many people in the area as possible. In this case,
the advertising agency or individual client may ask for TSA/NSI numbers only, disregarding the DMA and
metro data.

The cume (cumulative audience) or reach is an estimate of the number of different people who listened
to or viewed at least five minutes within a given daypart (the five minutes do not have to be
consecutive). The cume is also referred to as the unduplicated audience. For example, a person who
watches a soap opera at least five minutes each day Monday through Friday would be counted only
once in a cume rating, whereas the person’s viewing would be “duplicated” five times in determining
average quarterhours.

The average quarter-hour (AQH) is an estimated average of the number of persons or households tuned
to a specific station for at least five minutes during a 15-minute time segment (the five minutes do not
have to be consecutive). Unlike the cume, where a person is counted only once during a Monday– Friday
program, the listener or viewer would be counted five times for the time period. Cume represents the
number of different people in the audience; AQH represents the average number of people in the
audience.

Cume and AQH estimates are provided for the TSA/NSI, DMA, and MSA in all ratings books. Stations are
obviously interested in obtaining high AQH figures in all demographic areas because these figures
indicate how long an audience is tuned in and thus how loyal the audience is to the station. The AQH
data are used to determine the average radio listener’s time spent listening (TSL) or time spent viewing
(TSV) for television during a given day or daypart. All stations try to increase their audience TSL because
it means that the audience is not continually switching to other stations.

The gross rating points (GRPs) are a total of a station’s ratings during two or more dayparts and
estimate the size of the gross audience. Advertising purchases are often made via GRPs. For example, a
radio advertiser who purchases 10 commercials on a station may wish to know the gross audience that
will be reached. The gross rating point indicates that about % of the listening audience will be exposed
to the 10 commercials.

Decline in Ratings and Shares

The definition of success varies from one area to another. In broadcasting, the number of viewers or
listeners determines success— the networks or stations with the highest numbers (ratings and/or
shares) are considered successful—the winners. But, how many people are these successful radio and
television stations reaching? How big is the audience?

Until a few years ago, successful TV stations garnered ratings of 20–30% and shares of 30–40% for most
prime-time programs and nightly network news; successful radio stations had shares of 15–40% in most
daytime dayparts. This is no longer the case.

In television, it is now common to see the leading show with a 6.0 rating and 14 share. In radio, a
successful station garners a 3.0–5.0 share of the audience. The huge radio and TV audiences are gone;
they have disappeared because (a) there are many choices for listeners and viewers and (b) there is a
lack of concern by many broadcasters to find out what the audience wants to hear or see.

Have the lower numbers created a new atmosphere for broadcasters? What do broadcasters do about
the lower numbers? To be honest, most broadcasters “take it easy” when they reach the number one
position in their market (or the nation)— they perceive that they have given the audience what it wants.
Is that true? Let’s look at the situation from a different perspective. In radio, for example, a program
director and the morning show team may be ecstatic if they win the 6:00–10:00 a.m. daypart with a 4.0
share in the target demographic. Ecstatic! Bonuses are handed out, parties are held, and everyone is
happy (“We killed ’em.”). But wait a minute—the 4.0 share means that 96% of the target demographic is
not listening to their radio station! Is the radio station giving the audience what it wants to hear? Sure it
is—to 4% of them. This is the reality of today’s broadcasting.

Non Rating Research:

Although audience ratings are the most visible research data used in broadcasting, broadcasters,
production companies, advertisers, and broadcast consultants use numerous other methodologies.
Ratings provide estimates of audience size and composition. Nonratings research provides information
about what the audience likes and dislikes, analyses of different types of programming, demographic
and lifestyle information about the audience, and more. These data provide decision makers with
information they can use to eliminate some of the guesswork involved in giving the audience what it
wants. Nonratings research is important to broadcasters in all markets, and one characteristic of all
successful broadcast or cable operations is that the management uses research in all types of decision
making.

What is the importance of nonratings research to a newcomer to the broadcast field? Renowned
American broadcaster Frank Bell (2009) says:

Local market research provides something unattainable from inside a radio or TV station: the
unvarnished perspective of those wonderful people who actually tune in every week and keep us in
business. As a wise man said many years ago, “The only reality that counts is that of the audience.”

The following section describes some nonratings research methods conducted in the electronic media to
provide information to decision makers so that they don’t “fly blind,” as Frank Bell says.

Program Testing
Research is an accepted step in the development and production of programs and commercials. It is
now common to test these products in each state of development: initial idea or plan, rough cut, and
postproduction. A variety of research approaches can be used in each stage, depending on the purpose
of the study, the amount of time allowed for testing, and the types of decisions that will be made with
the results. The researcher must determine what information the decision makers will need to know and
must design an analysis to provide that information.

Since programs and commercials are expensive to produce, success-minded producers and directors are
interested in gathering preliminary reactions to a planned project. It would be ludicrous to spend
thousands or millions of dollars on a project that has no audience appeal.

Although the major TV networks, large advertising agencies, and production companies conduct most
program testing, many local TV stations are involved in programming research. Stations test promotional
campaigns, prime-time access scheduling, the acceptability of commercials, and various programming
strategies.

One way to collect preliminary data is to have respondents read a short statement that summarizes a
program or commercial and ask them for their opinions about the idea, their willingness to watch the
program, or their intent to buy the product based on the description. The results may provide an
indication of the potential success of a program or commercial.

However, program or commercial descriptions often cannot adequately describe the characters and
their relationships to other characters in the program or commercial. This can be done only through the
program dialogue and the characters’ on-screen performance.

If an idea tests well in the preliminary stages (or if the producer or advertiser wishes to go ahead with
the project regardless of what the research indicates), a model or simulation is produced. These media
“hardware” items are referred to as rough cuts, storyboards, photomatics, animatics, or executions.

The rough cut is a simplistic production that usually uses amateur actors, little or no editing, and
makeshift sets. The other models are photographs, pictures, or drawings of major scenes designed to
give the basic idea of a program or commercial to anyone who looks at them.

Testing rough cuts is not expensive, which is important if the tests show a lack of acceptance or
understanding of the product. The tests provide information about the script, characterizations,
character relationships, settings, cinematic approach, and overall appeal. Rough cut tests seldom
identify the reasons why a program or commercial tests poorly, but they provide an overall indication
that something is wrong and provide information for more specific tests.

When the final product is available, postproduction research can be conducted. Finished products are
tested in experimental theaters, in shopping centers (where mobile vans are used to show commercials
or programs), at respondents’ homes in cities where cable systems provide test channels, or via
telephone, in the case of radio commercials. Results from postproduction research often indicate that,
for example, the ending of a program is unacceptable and must be re-edited or reshot. Many problems
that were not foreseen during production may be encountered in postproduction research, and the data
usually provide producers with an initial audience reaction to the finished, or partially finished, product.

The major TV networks use their own approaches to testing new programs. One approach is to test pilot
programs on cable TV outlets throughout the country, where respondents are pre-recruited to watch
the program. Another approach is to test pilot programs in large focus group settings. Regardless of the
type of pretesting, the networks continually test the programs using a variety of qualitative and
quantitative approaches such as focus groups and telephone interviews.

There are many ways to test commercials or programs. Some researchers test commercials and
consumer products by showing different versions of commercials on cable systems. Prototype (test)
commercials can be “cut into” a program (they replace a regularly scheduled commercial with a test
spot) in target households. The other households on the cable system view the regular spot. Shortly
after airing the test commercial to the target households, follow-up research is conducted to determine
the success of the commercial or the response to a new consumer product.

Commercials can also be tested in focus groups, shopping center intercepts, and auditorium-type
situations. Leading companies and advertising agencies rarely show their commercials on television until
they are tested in a variety of situations. The companies and their agencies do not want to communicate
the wrong message to the audience.

Music Research
Music is the product of a music radio station, and failing to analyze the product invites ratings disaster.
To provide the radio station’s listeners with music they like to hear and to avoid the songs that they do
not like or are tired of hearing (burned out), radio programmers use two primary research procedures:
auditorium music testing and callout research.

Auditorium tests are designed to evaluate popular songs of the past few years and songs older than 10
years. New music releases can’t be tested satisfactorily in the auditorium or with callout procedures
because it is necessary to play the entire song. New music is frequently tested on the air during
programs with titles such as Smash or Trash, where listeners call in and voice their opinion about new
releases. Sometimes new music is tested in focus group situations where the respondents can listen to
the entire song.

Auditorium tests and callout research used for the same principle: to provide a program director or
music director with information about the songs the viewers likes, dislikes, is tired of hearing, or is
unfamiliar with. This information allows the program director to make decisions based on audience
reaction rather than gut feelings or guessing.
Research companies and program directors have a range of scales for listeners to use in evaluating the
music they hear. For an example, respondents can be asked to rate a hook on a 5, 7, or 10 point scale,
where 1 represents “hate” and 5, 7, or 10 represents “like a lot” or “favorite.” Or the scales can be used
without labels, and respondents are asked to rate the songs on the scale, where the higher the number,
the more the song is liked. There are also options for “unfamiliar” and “tired of hearing.”
Which scale is the best? Research conducted over several years by the senior author of this book
indicates that either a 7 point or 10 point scale provides the most reliable results.

Sometimes researchers inquire respondents to rate whether each song “fits” the music they hear on
their favorite radio station. This additional question helps program directors determine which of the
tested songs might not be appropriate for their station. In addition, some research companies inquire
listeners whether they would like radio stations in the area to play a particular song more, less, or the
same amount as they currently do. This is an inefficient and inaccurate way to determine the frequency
with which a song should be played. The reason is that there is no common definition of more, less, or
same and listeners are poor judges of how often a station currently plays the songs.
The bottom line in all music testing is that program directors should use the data as a guide for selecting
their radio station’s music, not as a music selection “bible.”

Auditorium Testing: In this method, between 75 and 200 people are invited to a large room or hall,
often a hotel ballroom. Subjects all the time match specific screening requirements determined by the
radio station or the research company, such as listeners between the ages of 25 and 40 who listen to
soft rock stations in the client’s market. Respondents are generally recruited by a field service that
specializes in recruiting people for focus groups or other similar research projects. Respondents are
generally paid between $25 and $150 for their cooperation. The auditorium setting—usually a
comfortable location away from distractions at home—allows researchers to test several hundred hooks
in one 90- to 120-minute session. Usually 200 to 600 hooks are tested, although some companies
routinely test up to 800 hooks in a single session. However, after 600 songs subject fatigue becomes
evident by explicit physical behavior (looking around the room, fidgeting, talking to neighbors), and
statistical reliability decreases. It is easy to demonstrate that scores for songs after 600 are not reliable
(Wimmer, 2001), specifically in reference to unstable standard deviations for the songs.

While auditorium music testing is a relatively simple project, there are many things to consider ensuring
that the test is successful. Among numerous procedures and steps, some basic procedures to follow
when conducting an auditorium music test include the following:

1. The key to a successful test is a good introduction that explains the purpose of the test and how
important the respondents’ answers are to hearing good music on the radio. It is important to
stress that there are no right or wrong answers in the test and that the goal is to collect a variety
of opinions.
2. The moderator must be in total control over the situation to ensure that respondents do not talk
amongst themselves or try to influence other respondents’ answers.
3. Adequate breaks must be taken during the session. Respondents shouldn’t listen to more than
200 songs without a break.
4. The moderator must make sure that all respondents understand the scoring system. After the
test begins, the moderator should check to see that each person is rating the songs correctly.
5. The moderator should not allow the respondents to sing along with the songs. This disrupts the
other respondents.
 6. The moderator must always expect the unexpected, including such things as electrical outages,
sick respondents, unruly respondents, equipment failures, or problems with the hotel
arrangements.

Callout Research: The idea of callout research is the same as that of auditorium testing; only the process
for collecting the data is changed. Instead of people being invited to a large hall or ballroom, randomly
selected or prerecruited subjects are called on the telephone. Subjects are given the same rating
instructions as in the auditorium test; they listen to the hook and provide a verbal response to the
researcher making the telephone call. Callout research is used to test only newer music releases. (The
reliability and validity of using the telephone for callout research is better than using the telephone for
auditorium testing because only a small number of songs are tested in callout.)

While callout methodology is adequate because only a few songs are tested, the limitation on the songs
tested is also the methodology’s major fault. Well-designed callout research involves testing a maximum
of 20 songs because subject fatigue sets in quickly over the telephone. Other problems include the
distractions that are often present in the home, the poor quality of sound transmission created by the
telephone equipment, and the fact that there is no way to determine exactly who is answering the
questions.
Miscellaneous Research
The electronic media are unique, and each requires a different type of research. Here are examples of
research conducted by and for stations:

1. Market studies. A market study investigates the opinions and perceptions of the entire market,
regularly within a specific age range, such as 25–44. There are no requirements for respondents
to meet in terms of station listening or viewing, and the sample matches the population
distribution and makeup of the market.

2. Format studies. A format study for a radio station involves a sample of respondents who listen
to certain type of music. These respondents are asked a sequence of questions to determine
which stations provide the best service in a variety of areas, such as music, news, traffic reports,
and community activities.
3. Format search studies. The title of the study explains its purpose, to find an available radio
format in a given market. An experienced researcher can precisely predict a potential format
hole with a specifically designed three-module questionnaire.

4. Program element importance. A program element importance study identifies the precise
elements on radio or television that are most important to a specific audience. Station managers
use this information to ensure that they are providing what the viewers want.

5. Station image. It is important for a station’s management to know how the audience perceives
the station and its services. Audience misperception of management’s purpose can decrease
audience size and, consequently, advertising revenue. For an example, assume a radio station
has been CHR (Contemporary Hits Radio) for 10 years and switches to a Country format. It is
important that the audience and advertisers are aware of this change and have a chance to say
their opinions. This can be accomplished through a station image study, where respondents are
asked questions such as “What type of music does FM play?” “What types of people do you
think listen to FM?” and “Did you know that FM now plays Country music?” If research tells that
only a few people are aware of the change in format, management can develop a new
promotional strategy. Station image studies are conducted every so often by most large radio
stations to gather current information on how the audience perceives each station in the
 market. If station managers are to provide the services that listeners and viewers want, they
must understand viewers trends and social changes.

6. Personality studies. Radio and television managers of successful stations regularly test the on-air
personalities. Announcers, news anchors, and all other personalities are tested for their overall
appeal and fit with other station personalities. Personality studies are frequently conducted for
stations to find new talent from other markets or even to test personalities who are on other
stations in the market with the intent of hiring them in the future.

7. Advertise analysis. To boost the value of their service to advertisers, many stations conduct
studies with local business executives. Some classic questions are “When did your business
open?” “How many natives own this business?” “How much do you invest in advertising per
year?” and “What exactly do you expect from your advertising?” Information gathered from
client questionnaire is used to help write more effective advertising copy, to develop better
advertising proposals, and to allow the sales staff to know more about each client. Normally, the
questionnaires are administered before a business becomes an advertiser on the station, but
they can also be conducted with advertisers who have done business with the station for several
years

8. Account executive research. Radio and television station managers all through the country
conduct surveys of advertising agency personnel, usually buyers, to determine how their sales
executives are perceived. It is crucially important to know how the buyers perceive the
salespeople. The outcome of the survey indicates which salespeople are performing well and
which may need additional help. These surveys frequently disclose that problems between a
sales executive and a buyer are due to personality differences, and the station can easily correct
the problem by assigning another salesperson to the business or advertising agency.

9. Sales research. In an effort to boost sales, many stations conduct research for local clients. For
an example, a station may conduct a “bank image” study of all banks in the area to determine
how residents perceive each bank and the service it provides. The outcome from such a study is
then used in an advertising proposal for the banks in the area. For an example, if it is discovered
 that First National Bank’s 24-hour automatic teller service is not well understood by local
residents, the station might develop an advertising proposal to concentrate on this point.

10. Diversification analyses. The goals of any business are to get bigger and to achieve higher
profits. In an effort to reach these goals, most of the larger stations, partnerships, and
companies engage in a variety of studies to determine where investments should be made.
Should other stations be bought? What extra types of activities should the business invest in?
Such type of studies is used for forecasting and represents a major portion of the research
undertaken by larger stations and companies. The changes in broadcast ownership rules made
by the FCC have considerably increased the amount of acquisition research conducted by
individuals, group owners, and other large companies in the broadcasting industry.

11. Qualitative research. Managers of successful broadcasting and cable operations leave nothing to
chance, which means that they test every part of their station or network. Research is organized
to test billboard advertising, logo designs, bumper stickers, bus advertising, direct mail
campaigns, programming interests, and more.

12. TV programming research. This is a wide category that includes testing local news programs,
promotional materials used by the station (known as topicals), entertainment programming, and
everything else that appears on the station.

5.1.3 Research in Advertising: Pre testing Methods: Direct Ratings, Portfolio Tests
Advertising research is to be conducted to measure the advertising impact or the result of the effort
with the help of thorough study on advertising objectives, product appeals, copy testing, and media
effectiveness.
Research takes place in advertising process at several points. Early on in the process, it is sometimes
used to help a marketer establish which segment of the market to target.
The history of advertising research explains us quite a bit about advertising research’s current status.
Although some advertising agencies have had research departments for more than 70 years, the real
boom days came between the 1930s and the 1970s.
Advertising is bringing a service to the attention of potential and current customers. Advertising is
paying attention on one particular product or service. As a result, an advertising plan for one product
might be very different than that for another product. It is normally done with signs, brochures,
commercials, direct mailings or e-mail messages, private contact, etc.
Endorsement keeps the product in the minds of the customer and helps stimulate demand for the
product. Endorsement involves ongoing advertising and publicity. The continuing actions of advertising,
sales and public relations are often considered aspects of promotions.

At a recent industry conference on account planning, it was actually strongly asserted that one of the
most beneficial forces of social justice on the scene today was a “new” form of advertising research-
account planning.
Research is a methodical and objective investigation of a subject or problem to find out relevant
information. Research might be fundamental or applied in nature.
Fundamental research looks to expand the boundaries of knowledge in a given area with no necessary
immediate application to existing problems. Applied research tries to use existing knowledge as an aid
to solve the given problem or set of problems.
Marketing researches are usually conducted by the companies for their internal use. The help of
marketing research firms may also be used for the benefit of the organization, and for analyzing
complicated problem oriented marketing situations.

It is the function of the advertiser to measure the effects of communication. The module of
communication model should be studied thoroughly and evaluation should be made on their
effectiveness.
The components of communications are as below
 Source factors
 Message variables
 Media strategies and
 Receivers of the message.
The advertising research is to be conducted to measure the advertising impact or the result of the effort
with the help of thorough study on advertising objectives, product appeals, copy testing, and media
effectiveness. The objectives of the research should focus on –
i. the optimum utilization of advertising budget
 ii. the choice of media in implementing an advertisement campaign
iii. the effect of advertisement on the target audience
iv. to bring cost-effectiveness in advertising.

Formulation of advertising objectives and funds of expenditures are to be considered in setting the goals
and the objectives should be streamlined in order to determine the tasks to be performed.

Significance of the advertisement is of great importance to create an impact and to venture the product
in the market.
End user’s attitude and the effect of message may be identified and measured. The advertiser should
locate most important characteristics of a particular market segment and should design the product
appeal and the relevant message accordingly. Rather than criticizing the competitor’s brand, the
advertiser should highlight the positive features of the brand advertised, through well directed message.

The techniques similar to depth interviews and projective tests may be conducted to estimate and
measure the spontaneous and emotional responses. A variety of methods of testing techniques may be
adopted to measure the effectiveness of the message.
The methods are as follows:
Pre testing Methods - Direct Ratings, Portfolio Tests
Before the advertisement is published for full run in a medium, these tests are conducted to make
modifications or improvements.
Portfolio tests

Portfolios of different different advertisements are prepared and are given to selected group of
individuals who are asked to go through this portfolio and can take as much time as they need. After
that portfolios are taken back and respondents are asked to recall the advertisements. The point of view
of the respondents is taken and conclusions drawn as to the extent that a message stands out or is
understood.

Even as portfolio tests offer the opportunity to compare alternative advertisements, directly, a number
of weaknesses limit their applicability.

i. Recall might not be the best test. Some researchers have difference of opinion that for certain
types of products—those of low involvement—ability to recognize the advertisement when
shown may be a better measure than recall.
ii. Factors apart from advertising creativity and presentation may affect recall. Curiosity in the
product or product category, the fact that respondents know they are participating in a test or
interviewer instructions may account for more differences than the advertisement itself.

The three main types of portfolios are: working portfolios, display portfolios, and assessment portfolios.

Though the types are distinct in theory, they tend to overlap in practice.
In this technique a number of dummy advertisements are put in a folio along with the advertisement to
be tested. The views and sense on the advertisement copies shown are collected from selective
respondents.

Alongside the regular advertisements some dummy copies are kept in a folio. Then the consumer-
sample sees the folio. The consumer is then asked about what he has seen in each advertisement. The
advertisement providing minimum playback is considered the best. But afterward it is necessary to
observe whether the chosen advertisement is dummy or regular. If it is found dummy the actual one is
improved on the same lines.

These tests are named after the way in which the advertisements to be tested are combined. A group of
advertisements, usually a mixture of test ads and control advertisements is placed in a portfolio. Now
and then the advertisements are actually placed in dummy copies of news papers or magazines.
Respondents who play role of or called representative of the target audience are given the folio and
asked to go through it, reading whatever interests them and taking as much time as they want. After
completing this task, the respondents are asked to recall the advertisements that they can remember.
For each recalled advertisement, the respondent is asked to play back as much of the advertisement as
possible. This information is recorded verbatim. Additional questions may be asked about such things as
the believability of the claims in the advertisement, the general reaction to the advertisement, and
whether the respondent uses the product.

Frequently, the portfolio test is used to test the merits of two or more alternatives advertisements. In
such cases, an experimental design is used in which two or more sets of folios are prepared. The only
difference between folios is that one set contains test advertisement A, another set contains test
advertisement B, and so on. The non-test advertisements are the same in all folios and are positioned in
the same order. By using small matched samples and comparing recall and playback scores among the
various groups a winner is obtained.

In summary:
Unit – 5.2: Testing Methods

At the end of this unit, you will be able to understand:

 Physiological Tests: The Eye Camera Test

 Physiological Tests: The Pupillo-meter Test
 Physiological Tests: The Tachisto Scope Test.
 Consumer Jury Test
 GSR Test
 Post Testing Methods: Enquiry Test
 Post Testing Methods: Recall Test
 Post Testing Methods: Recognition Test
 Post Testing Methods: DAR Testing 
5.2.1 Physiological Tests: The Eye Camera Test
This is one of the most widely used tests in media research. One can measures the actual viewing
behavior of the respondents using it. This camera is fitted with high power lenses and is focused on the
eyes of the respondents. It records continuously the behavior of the eye as it moves horizontally and
vertically while the respondent is watching pictures on a screen. It helps in finding out where the
respondent looked at any moment, how the eye moved over the image, and how long it determined on
a particular point. The eye camera test is used to validate print ads, posters, TV ads etc.

5.2.2 Physiological Tests: The Pupillo-meter Test:

These techniques measure the change in the size of the pupil of the eye as the respondent watches or
just take a glance at the advertisement. Bigger or expanded pupil means stronger interest and smaller or
slender pupil means lower interest. Every one of these mechanical devices used for laboratory testing
can only measure how respondents react to an advertisement while watching it. These devices cannot
measure attitude changes that take place long after the first viewing. None of the studies predict the
respondents' buying behavior. Another difficulty with laboratory tests using mechanical devices is that
such tests involve very small samples of the target market. Then there is the problem of the alien and
artificial environment of the laboratory, which is entirely different from real life situations.

5.2.3 Physiological Tests: The Tachisto Scope Test.

This is basically a slide projector, which can be operated under varying circumstances of speed and
illumination. This device helps find out how fast or slow the respondents perceive and respond to the
advertising message. Although the tachistoscope helps measure only the initial response after watching
the ads for the first time and not the response to ads after multiple viewing.

5.2.4 Consumer Jury Test

In this test a representative cross section of the target audience is selected as jury members to judge the
advertisements which are being tested. These people are most likely to be exposed to the final
advertisement. The response of this kind of jury is totally different from those of advertising personnel
who, on one hand, over estimate their knowledge of prospective buyers, while on the other hand, are
too concerned with technical aspects like copy, visuals, layout, filming etc. Respondents in a consumer
jury test are asked to evaluate one or more ads by some special criteria as attention getting power,
believability of claims etc. They are asked to rank the ads in the order of merit. To generate real and life-
like situations, print ads are inserted in magazines or TV ads are shown with some recorded programme.
Such tests usually help separate very weak ads from very strong advertisements. Also these tests do not
involve much time or money. However, such studies have some disadvantages too. Rating of the
respondents might be based on unimportant details. Consumer jury tests provide only the mindful
evaluation and not intuitive evaluation, which plays an important role in buying behavior. Also as the
respondents frequently play the role of experts they provide subjective evaluations, and not objective
evaluations based on their own reactions. And such tests are mostly confined to the notice ability of the
ads while ignoring other aspects. Also such studies do not provide much control as everything is in the
hands of the respondent.

5.2.6 Post Testing Methods: Enquiry Test

Enquiry Tests: This is a technique used in advertising research where advertising, brands, or products
are tested through random enquiries of respondents. The information acquired is usually used to
evaluate the various forms of advertising media. In some cases, the respondents are actually citizens
who have written unwanted comments to the advertiser. In different cases, the response has been
stimulated by an advertiser’s offer. Test of the effectiveness of advertising based on the number of
enquiries about an advertisement.
Some end user durables companies issue coupons as a part of the advertisement copy and when they
are circulated to the customers, they are supposed to fill it up and send it back to the company. So when
the customers are filling the coupons they are seeing the ad copy as well. So from the number of
coupons acknowledged estimation can be made as to the number of the readership of the
advertisement.

5.2.7 Post Testing Methods: Recall Test

Recall Test: The recall test is a sort of a memory test which attempts to determine the impression or
penetration made by an advertisement on the reader’s mind. It evaluates the memo ability of an
advertisement by contacting addressees and asking what they commit to memory about it. As a test of
memory, recall is more difficult than recognition. Various recall tests differ in the subject matter they
deal with and the aid-to-memory they provide. Some deal with print ads while others deal with
broadcast ads. Some recall tests use aids to help audience members to remember while others do not
use any such aids. These tests try to answer questions like:
 How many people recall seeing or hearing the ad?
  How much and what do they remember about it?
Some other methods deal with campaign themes and slogans and answer questions like: How fast and
how widely is the theme (idea) registering? Some other tests try to evaluate advertisements before they
are placed in the media and answer the question: How well are the 'test' advertisements recalled after
controlled exposures in dummy magazines or while being shown along a set of different ads with
recorded programmes? Other recall test methods try to find out what people have learned about the
product, the service or the company advertised. The aid-to-memory or the amount of help provided to
the respondents ranges from very little clues like the product category to a complete series of clues
designed to draw out the very last detail of the advertisement. According to the presence or absence of
clues or aids, recall tests are divided into two types - aided recall and unaided recall. Whatever the
method used, the final objective of recall tests is to find out whether the respondent has established a
strong connection between the brand name and the sales message.

5.2.8 Post Testing Methods: Recognition Test

Recognition Test: First mission of an advertisement is to be seen, read, or heard. Each advertisement
uses some means or other to get interest and hold it. Here comes the first giant hurdle - recognition. It is
just a matter of identifying an advertisement that one has seen before. Recognition is an essential state
for effective advertising. If an advertisement can’t pass this barrier, it will probably not be effective.
Recognition tests are generally used for print ads. As many types of recognition tests have been
designed, Daniel Starch of StarchINRA-Hooper Company developed the most widely used of such tests.
The Starch Recognition Test sends newspaper or magazines to respondents and then sends interviewers
to carry out the tests. Starch tests usually find out the 'recognition rates' of various elements of the ads
like the visual or illustration, headline, logo, size, shape etc. Such tests provide information regarding
 The percentage of readers, who remember seeing the advertisement (Noted),
 The percentage of readers who recall seeing or reading any part of the advertisement
identifying the product or brand, the company name, or logo (Associated),
 The percentage of readers who reported reading at least one part of the advertisement (Read
most).
One big advantage of the Starch Test is that it is simpler as it uses easy questions and can be conducted
quickly. Also Starch test has evolved over the last few decades into a highly precise method that covers
many product categories. The Starch test has proved highly reliable too. For television advertisements,
there is the Bruzzone-Test developed by the Bruzzone Research Company. This test is conducted
through mail. Respondents are sent questionnaires that show scenes from TV ads with the script, but
without the brand name. Respondents are asked whether they remember having seen the commercials
before. If the answer is 'yes', then they are asked to identify the brand. This is followed by a ranking of
the ads by the respondents on the basis of a list of adjectives. In addition to measuring recognition of TV
ads, this test also provides assessment of liking, preference and purchase etc. of the brands. Another
company that has developed Recognition tests for TV and radio ads is the Communis Company.
Communis tests involve showing brief and edited (5-10 seconds) TV ads without the brand and company
identification. The respondents are then asked to indicate if they have seen the ad, identify the brand
and point out elements of the ads, which have been edited out.
The advantages of recognition tests (particularly the Starch test) include that these tests answer the
following questions for an advertiser:
 To what extent is my advertisement seen or/and read?
 Are my present advertisements better read or watched than my previous ads?
 Are my advertisements better read than those of my competitors?
 Is the reading of the ads of my current campaign increasing or decreasing?
 How can I present my sales message that it will be better read?
The utmost benefit of the recognition tests is that it measures what an advertisement has achieved
under normal (real) conditions. Whilst response from the audience in terms of recognition does not
necessarily bring about sales, the achievement of attention and interest by an advertisement is an
important step that often leads to purchase. Another important aspect of the recognition test is its
ability to measure effectiveness of ads in all kinds of media and both in real and laboratory situations. In
addition, being a low cost method, it is used by most advertising researchers. Recognition tests,
however, are not final measures of advertising effectiveness. In fact, recognition is very easy to achieve
by using attention-getting devices like catchy visuals, unusual headlines, bizarre or odd shapes, flashy
and vivid color etc. And recognition alone does not ensure purchase by consumers. So advertisers
conduct other types of tests also in addition to recognition tests.

5.2.9 Post Testing Methods: DAR Testing 

DAR Testing: The Dar test or the day-after-recall test is a common method to evaluate television
commercials. Renowned researcher George Gallup first devised this in the early 1940's. Then he was
working with the Young and Rubicam advertising agency. After that he joined the Burke Marketing
Research Company, where he further developed this method. The Gallup Recall method involves
telephoning 150 to 300 programme spectators 24 hours after a particular television commercial is
broadcasted. Some of the companies use a different time period, such as 48 hours or 72 hours.
Respondents are asked if they can recall any commercial telecast the previous day in a product category
(which the advertised brand belongs to). This is independently recall, as the brand name or company
name not mentioned. If they can’t recall the advertisement being tested with no clue, then they are
given clues about it. This is aided recall. Once recall of the advertisement happens, they are asked what
they learn about the advertisement, what was thought, what was displayed, and what was the main
idea. They are asked about the sales message as well, the story line, the plot, some visual or audio
elements so on. After the interview over telephone is over, then the 'scores' are measured in terms of
the following:
 Proved Name Registration (PNR) score, which is the % of audience members who recalled the
advertisement and proved recall by describing the advertisement accurately.
 An Idea Playback Profile that measures the level of sales message recall.
 A Favorable Buying Attitude scores those measures its persuasiveness. These 'scores' are
compared with proved norms or average scores of ads of similar length, similar product
categories, and similar brands.
While the DAR test is an 'on air' test in which only those ads are evaluated which have been telecast
and seen by people in real, home situations. At times, the ad to be tested is shown on a local cable
channel and viewers are told in advance to watch the programme on that channel. This method
saves a lot of money, time and efforts. In addition to the Gallup Recall Test for TV ads, there are
other methods also. The Gallup-Robinson Test is specifically designed for print ads. Mapes and Ross
Inc. tests recall for both print and broadcast ads. The main benefit of recall tests is based on the idea
that if people recall the ad (or any part of it) then they may buy it. This is because it is thought that
top-of-the-mind-recall plays a crucial role in the purchase behavior. But the criticism is based on the
argument that do people buy a brand if they remember the same?

Unit – 5.3: Statistical Application in Media Research

At the end of this unit, you will be able to understand:

 Statistical Application in Media Research ; Statistics Applied in Research, Frequencies and

Percentages
 Measures of Central Tendency: Mean, Median and Mode;
 Measures of Dispersion: Range, Standard Deviation and Mean Deviation;
 Association & Simple Correlation.

Introduction

Media Research generates great amount of data. Significant information is to be drawn from this data. A
statistical method helps in this process. Statistics are mathematical methods to collect, organize,
summarize, and analyze data. Statistics cannot perform miracles. Statistics alone will not “correct” a
misdirected, poorly phrased, or ambiguous research question or hypothesis, or a study that uses sloppy
measurement and design and contains numerous errors. Statistics provide valid and reliable results only
when the data collection and research methods follow established scientific procedures.

The science of statistics and the ease with which they can be used have changed dramatically since the
development of computers. Only a few decades ago, researchers spent weeks or months performing
hand- calculated statistical procedures— calculations that now take only seconds or minutes on a hand-
held calculator or personal computer. In addition, there are now dozens of excellent computer software
programs and spreadsheets to calculate nearly any type of statistical test, all of which have simplified all
types of statistical analysis.

5.3.1 Statistical Application in Media Research; Statistics Applied in Research, Frequencies and
Percentages

Statistical Methods: Quantitative data show certain general characteristics, which can be briefly put
together as follows:

 They show a tendency to focus at certain values, generally somewhere in the centre of the
distribution. Measures of this tendency are called measures of central tendency or averages.
 The data fluctuate about a measure of central tendency. This refers to the inconsistency
characteristics of the data, often called dispersion.

Descriptive statistics condense data sets to allow for easier interpretation. If a random sample of 100
people were asked how long they listened to the radio yesterday and all 100 answers were recorded on
a sheet of paper, it would be difficult to draw conclusions by simply looking at that list. Analysis of this
information is much easier if the data are organized in some meaningful way—the function of
descriptive statistics. These statistical methods allow researchers to take random data and organize
them into some type of order.

During a research study, investigators typically collect data consisting of measurements or observations
of the people or items in a sample. These data usually have little meaning or usefulness until they are
displayed or summarized using one of the techniques of descriptive statistics. Mass media researchers
use two primary methods to make their data more manageable: data distributions and summary
statistics.

Data Distributions A data distribution is simply a collection of numbers. Table 1 shows a hypothetical
distribution of 20 respondents’ answers to the question, “How many hours did you spend in the past
two days listening to the radio and watching TV?” While the distribution may appear adequate, it is
difficult to draw any conclusions or to generalize from this collection of unordered scores.

As a preliminary step toward making these numbers more manageable, the data may be arranged in a
frequency distribution—a table of the scores ordered according to magnitude and frequency of
occurrence. Table 2 presents the data from the hypothetical radio/TV survey in a frequency distribution.

Table: 1 Distribution of Responses to “How many hours did you spend in the past two days listening to
the radio and watching TV?”
Respondent Hours
A 12
B 9
C 18
D 8
E 19
F 21
G 15
H 8
I 11
J 6
Respondent Hours
K 14
L 16
M 23
N 25
O 11
P 14
Q 12
R 19
S 21
T 11
Table 2 Frequency Distribution of Responses to “How many hours did you spend in the last two days
listening to the radio and watching TV?”
Hours Frequency (N=20)
6 1
8 2
9 1
11 3
12 2
14 2
15 1
16 1
18 1
19 2
21 2
23 1
25 1

Now the data begin to show a pattern. Note that the typical frequency distribution table includes two
columns. The column on the left contains all the values of the variable under study; the column on the
right shows the number of occurrences of each value. The sum of the frequency column is the number
(N) of persons or items in the distribution

A frequency distribution can also be constructed using grouped intervals, each of which contain several
score levels. Table 3 shows the data from the hypothetical survey with the scores grouped together in
intervals. This table is a more compact frequency distribution than Table 2, but the scores have lost their
individual identity.

Table 10.3 Frequency Distribution of Radio and TV Listening and Viewing Hours Grouped in Intervals
Hours Frequency
0-10 4
11-15 4
 16-20 8
21-25 4

Other columns can be included in frequency distribution tables, such as proportions or percentages. To
obtain the percentage of a response, divide the frequency of the individual responses by N, the total
number of responses in the distribution. Percentages allow comparisons to be made between different
frequency distributions that are based on different values of N.

Some frequency distributions include the cumulative frequency (cf), constructed by adding the number
of scores in one interval to the number of scores in the intervals above it. Table 4 displays the frequency
distribution from Table 2 with the addition of a percentage column, a cumulative frequency column, and
a column showing cumulative frequency as a percentage of N.

Table 4 Frequency Distribution with Columns for Percentage, Cumulative Frequency, and Cumulative
Frequency as a Percentage of N
Hours Frequency Percentage cf cf Percentage of N
6 1 5 1 5
8 2 10 3 15
9 1 5 4 20
11 3 15 7 35
12 2 10 9 45
14 2 10 11 55
15 1 5 12 60
16 1 5 13 65
18 1 5 14 70
19 2 10 16 80
21 2 10 18 90
23 1 5 19 95
25 1 5 20 100
N=20 100%
Sometimes it is best to present data in graph form. The graphs shown in this section contain the same
information as frequency distributions. Graphs usually consist of two perpendicular lines, the x-axis, or
abscissa (horizontal), and the y-axis, or ordinate (vertical). Over the years, statisticians have developed
certain standards regarding graphic format. One common standard is to list the scores along the x-axis
and the frequency or relative frequency along the y-axis. Thus, the height of a line or bar indicates the
frequency of a score. One common form of graph is the histogram, or bar chart, where vertical bars
represent frequencies. Figure 10.1 is a histogram constructed from the data in Table 10.1. Note that the
scores on the x-axis are actually the scores (hours) listed from the lowest value to the highest; the y-axis
shows the frequency of scores.
If a line is drawn from the midpoint of each interval at its peak along the y-axis to each adjacent
midpoint/peak, the resulting graph is called a frequency polygon. Figure 2 shows a frequency polygon
superimposed on the histogram from Figure 1. As can be seen, the two figures display the same
information.
A frequency curve is similar to a frequency polygon except that the points are connected by a
continuous, unbroken curve instead of by lines. The curve assumes that any irregularities shown in a
frequency polygon are simply due to chance and that the variable being studied is distributed
continuously over the population. Figure 3 superimposes a frequency curve onto the frequency polygon
shown in Figure 2.
Frequency curves are described in relation to the normal curve, a symmetrical bellshaped curve whose
properties are discussed more fully later in this chapter. Figure 10.4 illustrates the normal curve and
shows the ways in which a frequency curve can deviate from it. These patterns of deviation are called
skewness.

Skewness refers to the concentration of scores around a particular point on the x-axis. If this
concentration lies toward the low end of the scale, with the tail of the curve trailing off to the right, the
curve is called a right skew. Conversely, if the tail of the curve trails off to the left, it is a left skew. If the
halves of the curve are identical, it is symmetrical, or normal.
A normal distribution of data is not skewed in either direction. If data produce a curve that deviates
substantially from the normal curve, then the data may have to be transformed in some way to achieve
a more normal distribution.
5.3.2 Measures of Central Tendency

The main characteristics of a set of data is it’s measure of central tendency.

Take an example we may have data about the reading hours of the newspaper of different people in a
particular city. In place of dealing with reading hours data of individual a researcher may be paying
attention in ascertaining the average reading hours of the person in the city. Similarly in order to
ascertain the degree of power deficiency the administrator in-charge of power will be interested to
know the average daily power demand and average daily power generation in a state.

What is the significance of this term average? In general words, when we speak about average hours, or
average power shortage, we are really looking for a central figure amongst a number of observations,
which is representative of the entire group. Similarly when we say one is an average student what it
means is that he/she is typical of the group of which he/she is a part. Therefore, an average is a single
value, which is used to represent an entire set of data. This value permits us to judge against individual
values in the data with it and also permits us to compare different sets of data. The three principal
measures of central tendency extensively used in statistical analysis are the mean, median and mode.

Mean:
The mean, which is simply referred to as “arithmetic mean” is the most commonly used average. It is
addition of the values observed divided by the total number of observation of the data set. Assume x1,
x2, …………………, xn are the n number of observations which are contained in the sample.

Then there mean would be :

In calculating mean (arithmetic mean) with the help of this formula, there are two steps:
i. The observation of the data and take the sum of all obtain ∑Sx
ii. Then divide the total by the total number of observation i.e.n.

For example: Discover the arithmetic mean of hours of reading the newspaper per day of five
persons.
Therefore the average hours per person are .87 hours.
If the data is provided in the form of frequency distribution, then the mean x is computed as :

Mean Of Grouped Data:

If a frequency distribution of the data is provided in the forms of classes we no longer have the
original values of the observations. Let us have the following data on the earnings of 100 workers.
 Income frequency (f) m mf
20-40 6 30 180
40-60 9 50 450
60-80 11 70 770
80-100 14 90 1260
100-120 20 110 2200
120-140 15 130 1950
140-160 10 150 1500
160-180 8 170 1360
180-200 7 190 1330

In the above data, the authentic values of the observations are not given to us. Take the case of first
class, it shows there are 6 workers having income between Rs.20 to Rs.40. Actual income of these
workers is unknown to us. Therefore to find the mean of this type of data, we will have to deal with
the representatives of each class. Within each class, each item is assumed to have the value of the
midpoint of that class. The midpoint in every class has to be taken for each item in the class it
represents. The mid points computed from every class of the above data are shown by ‘m’. To
compute the mean of the data, now multiply the values of ‘m’ with corresponding frequencies and
then take the total of these products. Just as in the ungrouped data, we divide this total by the total
number of observations in the distribution to obtain the arithmetic mean. The method for
computing mean (arithmatic mean) is:

So average income of the workers is 110/-. For short-out method for computing arithmetic
mean we use the following formula:

Where Xd = assumed mean

 d= deviations of mid points from the assumed mean. Like, Xd

Median:
The median of group observations is defined as the middle most observation in the sense that one half
of the figure of observations is smaller than the median while the other half of observation is greater
than the median.
To determine the median, it is essential to first arrange the observation either in a descending or
ascending order of magnitude and then count their number.
If the number of observation is odd:
Median is equal to (n+1)/2 of the observation.
If the number is even:
Median is mean (arithmetic mean) of n/2 and n/2+1 the observations.

Take an example of this, the median of a series having five observations (6,9,10,11,13) is 10, 3 rd,
observation i.e. (5+1)/2 the observation. Had 15 been the sixth observations in the series, the median
would be (10+11)/2=10.5 i.e. arithmetic mean of – 6/2th and (6/2)+1 th observations.

Median is also known as positional average. The phrase position refers to the place of a value in a series.
Media is measured as a comparison of arithmetic mean when data consist acute values.
Take an example of below:
If the data have the following observations.
10, 13, 15, 17, 90.

The median of this series is 3rd observation which is 15. Now execute the arithmetic of this series, which
is equal to 29. It shows, mean is not representative of this series. The median M=15, on the other hand
is a representative measure of the observations.

When the data is provided in the form of frequency distribution, to find the median one can use the
following procedure:
 Enter the data in ascending or descending order of magnitude.
 Calculate the cumulative frequencies i.e. take the sum of successive frequencies corresponding
to given values of the variable.
  Locate the median which is = (n+1)/2
 From the column of cumulative frequency locate that which is equal to (N+1)/2 or next higher to
that and determine the value of the variable corresponding to it. That will be the value of
median.

Another example: Calculate the median by using the following data on the wages of the workers:

Wages 20 21 22 23 24 25 26 27 28
Number of workers 8 10 11 16 20 25 15 9 6

Answer: First put the values of the variable into ascending order

Wages Numbers of workers(f) Cumulative frequencies (cf)

20 8 8
21 10 18
22 11 29
23 16 45
24 20 65
25 25 90
26 15 105
27 9 114
28 6 120

Median (N + 1)/2 the Observation = (120+1)/2 = 60.5 th observation

This value falls in the cumulative frequency whose value is 65. Equivalent to this cumulative frequency
the value of our variable is 24.
Hence Median = 24

Median for grouped data: In group data, the individual observations have lost their individuality, and
the middle observation cannot be found by counting. It is required to get inside of a class to find the
value that divides the number of all observation in 1/2. If we divide the number of frequencies (N) in
two halves, then we find that the middle observation falls within a class. To position this class, we
cumulate frequency till we reach the lowest class whose cumulative frequency is greater than N/2, This
class is called the median class. After that, we locate the median value from the median class.
The method for this procedure is

Here:
e is the lower limit of median class
N is the total frequency.
fi is the sum of all frequencies accumulated before entering the median class.
ƒ med is the frequencies of the median class,
and
i is the width of the median class.

For example: Find the median of the following data.

Marks Numbers
0-4 10
4-8 12
8-12 18
12-14 7
14-18 5
18-20 8
20-25 4
25 and above 6

Solution: First of all put the data in ascending or descending out and after that find the median.
 Marks Number of students (f) Cumulative frequency (c.f)
0-4 10 10
4-8 12 22
8-12 18 40
12-14 7 47
14-18 5 52
18-20 8 60
20-25 4 64
25 and above 6 70

Median = size of (N/2)th item

= 70/2 = 35th observation.
Median exists in the class 8-12 = 10.88

Hence median is 10.88 marks.

Mode: In many situations mean (arithmetic mean) and median fail to reveal the true characteristics of
the data. When we talk about average consumer, average student, average size of shoes or ready-made
garments we have in mind mode and not the mean or median. In statistics the value of variable, which
occurs most often, is called mode.

Mode for ungrouped data: Mode basically does not require any calculation in case of ungrouped data.
All which needs to be done is to find the value that occurs most frequently.
For example, in a set of observation consisting of the values 8, 10, 12,12, 12, 16 modes is 12 because it
occurs twice as compared to other values, which occurred only once. Arranging the observations in an
array and counting the number of times each observation occurs can easily find out the mode. If the
data is given in the form of frequency distribution, mode can be identified just by looking to that value
of the variable, which has corresponding highest frequency.

For an example: Calculate the mode of the following data

Shoe size 5 7 8 9 10
Number of persons (frequency) 4 17 22 5 2

Answer: In the above example the value 8 has occurred the maximum number of times i.e. 22. Hence
the mode is 8, which is most common size of shoes.

Mode for grouped data: To calculate the mode of the grouped data we have to pass through he
following steps:
i. Identify the maximum class frequency.
ii. Identify the class corresponding to this frequency, which is called the modal class.
iii. Identify the class size using (upper limit – lower limit.)
iv. Calculate mode using the formula below:

Here l is the lower limit of modal class

h is the size of class interval

f is the frequency of the modal class.

f0 is the frequency of the class preceding the modal class.

f2 is the frequency of the class succeeding the modal class.

Example: Find the mode of below given data

 Class Frequency
50-55 2
55-60 7
60-65 8
65-70 4

Answer:

8 is the maximum class frequency

For this modal class is 60-65

Class size is h = 65-60 = 5

f1 = 8

f0 = 7

f2 = 4

l = 60

Hence mode = 60 + ((8-7)/ (16 – 7 – 4)]5

= 60 + (1/5)5

= 61
5.3.3 Measures of Dispersion

The average is a representative of the group of observation in the sense that it enables us to recognize
the centre. But it does not disclose how the individual observations are spread out on each side of the
centre. The information of variability of the individual observation around the average is extremely
important for testing the reliability of average.

Let’s suppose the average per capita income of a country is very high. On the basis of this it may be
concluded that living standard of people there is very high. If there is high variability in the incomes of
the citizens and income of the country is concentrated in the hands of few citizens, in this case our
conclusion based on average income may be wrong. Average is hardly fully representative of a group of
observation unless we know the manner in which the individual observations scatter around it in short a
further description of the data is essential if we want to know how representative the average is.

The measurement of the magnitude of spreads of the individual observations about the average is called
a measure of dispersion; it’s this measure that tells us how representative the average is as a description
of the data. If the observations spread in a very narrow range, then average would appear to be highly
representative.

There are five measures of dispersion – named the range, quartile deviation, mean deviation, variance
and standard deviation.

Range:

Range is the simplest potential measure of dispersion and is defined as the difference between the
highest and the lowest value in a set of data/ observation.
Hence, Range, R = (L-S),
Where L is largest value and S is smallest value in a set of data
The usefulness of range is that it gives an idea of the variability very quickly, but the disadvantage is that
range is affected greatly by fluctuations of sampling. Its value is by no means stable, being based on only
two values of the variable. As such, range is typically used as a rough measure of variability and is not
considered as a suitable measure in serious research studies.

Example: Find the range from below following observation

Class Frequency
50-55 2
55-60 7
60-65 8
65-70 4
Solution: Range of the data will be R = L-S = 70 – 50 = 20

Standard Deviation:
To calculate the standard deviation of the ungrouped data, first locate the deviation of each observation
of the data from the arithmetic mean of the data. Then the square these deviation and obtain the sum
of them. In the next step, divide the sum of square of deviation by the total number of observations and
then take the under root of it, to get the standard deviation.

We denote the population standard deviation by Greek letter a σ(sigma), while in case of sample it is
denoted by ‘s’.

Where σ is standard deviation of a sample,

X is each value in the data set,

X (bar) is mean of all values in data set,

N is number of values in the data set

Example: Calculate the SD for below ungrouped observation

3, 10, 9, 9, 4, 7, 14
Solution:

Mean (X (bar)) = (3+10+9+9+4+7+17)/7 = 8

N=7

Σ of square of (X – X(bar)) = 25+4+1+1+16+1+36 = 84

σ = √ 84/7 = √ 12 = 3.46

If the data is provided in the form of frequency distribution then below steps are taken to calculate the
SD:

i. Take the deviation of every observation from the mean and then square it
ii. Multiply these deviation by the respective frequencies and get the total
iii. Multiply this deviation by the respective frequencies and obtain the total. The above steps could
be summarized in the following formula

σ=
√ ∑ f .( X −X )2
N

Standard deviation of grouped data: In this case to find the standard deviation, we adopt the following
the procedure:

i. Find the middle points of each class.

ii. Get the deviation of these mid points from the mean of the data.
iii. Square the deviations and multiply them with the respective frequencies of each class and
obtain the total.
iv. Divide the total of above (iii) step by total number of observations and take the root of it to get.
The formula of above steps is

σ=
√ ∑ f .( X −X )2
N
Mean Deviation: Mean deviation of a series is the arithmetic average of the deviation of a range of
objects from a measure of central tendency (mean, median and mode). Mean deviation is based on the
objects of the distribution and is calculated as an average, on the basis of deviation obtained from
mean, median or mode but usually from the median. First we compute deviations of all the objects from
either mean or median ignoring plus (+) and (–) signs. They are known as absolute values of deviations
where the two parallel bars (ii) indicate that the absolute value is taken. It’s also called modules value.
Then the aggregate of these deviations are divided by the number of observations this is called mean
deviation.

Calculation of mean deviation

i. Arrange the data in ascending order

ii. Calculate median or mean or mode
iii. Get deviations of items from median/mean ignoring ± signs and denote the column as |D|
iv. Now calculate the sum of these deviation in case of discrete and continuous series |D| is
multiplied by respective frequency of the item to get Σf |D|
 v. Divide the total obtained by number of objects to get mean deviation
¿¿
M.D = Σf ∨D∨
N
vi. Apply the formula to get coefficient of mean deviation
Coefficient of M.D. = M.D /Median or Mean or Mode

Example: Calculate the mean deviation for below observation

Size of objects 6 12 18 24 30 36 42

Frequency 4 7 9 18 15 10 5

Solution:

X f cf |D| f|D|
6 4 4 18 72

12 7 11 12 84
18 9 20 6 54

24 18 38 0 0
30 15 53 6 90

36 10 63 12 120
42 5 68 18 90

Σf = 68 Σ|D| = 72 Σf|D| = 510

N + 1 th
Median = Size of ¿ ¿ item = 34.5th item
2

¿ ¿ 510
Median deviation = Σf ∨D∨
N = 68 = 7.5
5.3.4 Association & Simple Correlation.
Correlation analysis explores the association among two or more variables and makes inference about
the strength of the relationship.
Note: It is general to use the terms correlation and association interchangeably. Technically, association
refers to any connection between two variables, whereas correlation is often used to refer only to a
linear connection between two variables. The terms are used interchangeably in this guide, as is
common in most statistics texts.
Scatter plot
The scatter plot shows the association between two variables. A scatter plot matrix shows all pairwise
scatter plots for numerous variables.

Covariance
Covariance is a assess of how much two variables change together. A covariance matrix measures the
covariance between lots of pairs of variables.

Correlation coefficient
A correlation coefficient measures the connection between two variables. A correlation matrix measures
the correlation between a lot of pairs of variables.

Inferences about association

Inferences about the strength of association between variables are prepared using a random bivariate
sample of data drawn from the population of interest.

Basic Correlation Statistics

Suppose a researcher hypothesizes a connection between the number of pictures on the cover of a
magazine and the total number of copies sold. If the observations tell that more magazines are sold
when extra pictures are used, there may be a relationship between the two variables. Numerical
expressions of the scale to which two variables change in relation to each other are called measures of
association, or correlation. When making two different measurements of the same individual,
researchers commonly assign one measure as the X variable and the other as the Y variable. Like for an
example, in a study of whether a relationship exists between the size of a subject’s family and the
frequency with which that person reads a newspaper, the measure of family size could be the X variable
and the measure of newspaper reading the Y variable. Note that each subject in the collection under
study must be measured for both variables.

Below figure 1 shows hypothetical data collected from a study of eight subjects. Here the Y variable is
the number of times per week the newspaper is read, and the X variable is the number of persons in the
household. The two scores for each subject are plotted on a scatter gram, a graphic technique for
portraying a relationship between two or more variables. As indicated, family size and newspaper
reading enhance together. It’s an example of a positive relationship.

Negative (or inverse) relationship exists when one variable increases and the other correspondingly
decreases. Sometimes the relationship between two variables is positive up to a point and then
becomes negative or vice versa. At the point of time this happens, the relationship is said to be
curvilinear. When there is no trend for a high score on one variable to be associated with a high or low
score on another variable, the two are said to be uncorrelated. Figure 2 explains these relationships.

Numerous statistics are available to measure the level of relationship between two variables, but the
most frequently used is the Pearson product-moment correlation, symbolized as r. It varies in between
21.00 and 11.00. A correlation coefficient of 11.00 indicates a just right positive correlation: X and Y are
completely covariant (they vary together). A Pearson r of 21.00 indicates a perfect relationship in the
negative direction. The smallest value the Pearson r can comprise is 0.00, which represents absolutely
no relationship between two variables. Thus the Pearson r contains two types of information: (1) an
estimate of the strength of the relationship, as showed by the number, and (2) a statement about the
direction of the relationship, as shown by the 1 or 2 sign. Bear in mind that the strength of the
relationship depends solely on the number, and it must be interpreted in terms of absolute value. A
correlation of 2.84 is a stronger relationship than one of 1.24.

The formula for calculating r is:

Here X and Y stand for the original scores,

N is the number of pairs of scores, and
S again is the summation symbol.
The only new term is SXY, which stands for the sum of the products of each X and Y.
To find this quantity, just multiply each X variable by its corresponding Y variable and then add the
results. Table 1 demonstrates a computation of r.
A correlation coefficient is a pure number; it is not articulated in feet, inches, or pounds, nor is it a
proportion or percentage. The Pearson r is independent of the size and units of measurement of the
original data. Because of its theoretical nature, r must be interpreted with care. In particular, it is not as
easy as it sounds to conclude whether a correlation is large or small. Some writers have suggested a
variety of adjectives to describe certain ranges of r. For example, an r between .30 and .40 might be
called a “moderate” or “substantial” relationship, whereas an r of .61 to .80 might be termed “very
high.”
These labels are helpful, but they may lead to confusion. The best advice is to consider the nature of the
study. For example, an r of .60 between frequency of viewing television violence and frequency of arrest
for violent crimes would be more than substantial; it would be phenomenal. On the contrary, a
correlation of .60 between two coders’ timings of the lengths of news stories on the evening news is low
enough to call the reliability of the study into question. In addition, correlation does not in itself imply
causation. Newspaper reading and income might be strongly related, but this does not mean that
earning a high salary directs people to read the newspaper. Correlation is just one factor in determining
causality moreover; a large r does not necessarily mean that the two sets of correlated scores are equal.
The correlation figure tells nothing about the amount of time spent with each medium. It only suggests
that there is a sturdy likelihood that people who spend time reading newspapers also spend time
watching TV news.

Perhaps the best way to understand r is in terms of the coefficient of determination, or the proportion
of the total variation of one measure that can be determined by the other. It is calculated by squaring
the Pearson r to appear at a ratio of the two variances: The denominator of this ratio is the total
variance of one of the variables, and the numerator is the part of the total variance that can be
attributed to the other variable.
Take a look at the example, if r 5 .40, then r 2 5 .16. One variable explains 16% of the variation in the
other. Clearly, if r 5 1.00, then r 2 5 100%; one variable allows perfect predictability of the other. The
quantity 1 2 r 2 is usually called the coefficient of no determination since it represents that proportion
of the variance left unaccounted for or unexplained.

The Pearson r can be calculated between any two sets of scores. For the statistic to be a valid
description of the relationship, however, below three assumptions must be made:
 Data represent interval or ratio measurements
 the relationship between X and Y is linear, not curvilinear and
 The distributions of the X and Y variables are symmetrical and comparable.
If these assumptions can’t be made, the researcher must use another kind of correlation coefficient,
such as Spearman’s rho or Kendall’s W. For a thorough discussion of these and other correlation
coefficients, consult Nunnally (1994).

Partial Correlation
Partial correlation is a technique researcher’s uses when they believe that a confounding or spurious
variable may affect the relationship between the independent variables and the dependent variable: If
such a pressure is perceived, they can “partial out” or control the confounding variable.
For example, think about a study of the relationship between exposure to television commercials and
purchasing the advertised products. The researchers select two commercials for a liquid laundry
detergent (a “straight sell” version with no special video/audio effects, and a “hard sell” version that
includes special effects). The commercials are then shown to two groups of subjects: people who only
use powdered detergent and people who only use liquid detergent. The study design is displayed in
Figure 3.

If the outcome shows a very low correlation, indicating that any prediction made based on these two
variables would be very shaky, the researchers should investigate the presence of a confounding
variable. An assessment might reveal, for example, that the technicians had problems adjusting the
color definition of the recording equipment; instead of its natural blue color, the detergent appeared
dingy grey on the television screen. The study might be repeated to control (statistically eliminate) this
variable by filming new commercials with the color controls properly adjusted. The design for the new
study is displayed in Figure 4.
The partial correlation statistical procedure would facilitate the researchers to determine the influence
of the controlled variable. With the new statistical method, the correlation might increase from the
original study.