Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 10

CONTENT

 INTRODUCTION
 PROBLEMS IN PSYCHOLOGICAL MEASUREMENTS
1. Indirectness of Measurement
2. Lack of Absolute Zero
3. We Measure a Sample of Behaviour not the Complete Behaviour
4. Lack of Sufficient Stimulus/Responses Threshold
5. Uncertainty and Desirability Involved in Human Responses
6. Variability of Human Attributes Over Time
7. Problem of Quantification
 ERRORS OF MEASUREMENT
1.
Accidental/Chance Error
2.
Systematic/Biased Error
3.
Interpretative Error
4.
Variable Error
5.
Personal Error
6.
Constant Error
7.
Statistical Error
a) Errors of Descriptive Statistics
i. standard errors
ii. probable errors.
b) Inferential Errors
i. Type I error
ii. Type II error
 CONCLUSION

INTRODUCTION
 Psychological measurement seeks to validly measure differences between individuals
and groups in psychosocial qualities such as attitudes and personality.
 Psychological measurement focuses on developing and testing tools to validity
measure psychological qualities (such as knowledge, attitudes and opinions,
emotions, cognitions, and personality).
 Psychological tests (also known as mental measurements, psychological instruments,
psychometric tests, inventories, rating scales) are standardized measures of a
particular psychological variable such as personality, intelligence, or emotional
functioning.

PROBLEMS IN PSYCHOLOGICAL MEASUREMENTS


1. Indirectness of Measurement
Various psychological attributes are accessible to research and measurement only indirectly.
Generally, the attributes underlying psychological processes are presumed to be manifesting
themselves through overt behaviours only, which are considered as objective and
quantifiable. For example, if a researcher is interested in measuring the personality
dimensions of a subject, then it is something that is not directly available for measurement
as physical quantities—like length—are visible and concretely available for observation and
assessment. The only way to measure it is to assess the person on a set of overt or covert
responses (for example, by administering a psychological test) related to his personality or
other psychological attributes of interest. This renders the meaning and scope of
psychological measurement limited and inexact.

2. Lack of Absolute Zero


Absolute zero, in case of psychological measurement, means a situation where the property
being measured does not exist. The absolute zero is available in case of physical quantities,
like length, but is very difficult to decide in the case of psychological attributes. For example,
suppose an investigator wants to measure shyness, or say, attitude towards fashion, then it
is very difficult to define and find a situation where there is absence of shyness or the
attitude towards fashion.

3. We Measure a Sample of Behaviour not the Complete Behaviour


In psychological measurements, a complete set of behavioural dimensions is not possible
and we take only a carefully chosen sample of behavioural dimensions to assess the
attributes in question. For example, the Weschler Intelligence Scale uses carefully chosen 35
words from the English dictionary to judge the vocabulary of the test taker. Although the
sample is chosen only after fulfilling the various psychometric criteria, like randomness,
representativeness, and so on, it is always questionable to reach at a conclusion about an
aspect of behaviour, only by measuring a small, though representative part of it.

4. Lack of Sufficient Stimulus/Responses Threshold


Another problem encountered in psychological measurement is the creation of sufficient
amount of variable strength or threshold, which is actually relevant while studying the
particular psychological attribute. Experimental method, the main method behind the
scientificity of psychology, heavily suffers from this problem. Apart from this, it is also very
difficult to decide the relevant levels of the response threshold that is adequate for an
accurate prediction of the behavioural dimension in question.

5. Uncertainty and Desirability Involved in Human Responses


Test subjects often give uncertain and desirable responses which generally negates the
entire purpose of the psychological measurement. Uncertainty may arise either due to the
negligence on part of the researcher, or carelessness on part of the subject(s), or due to
uncontrolled extraneous variables. Although the researcher may correct such problems
arising out of uncertainties involved in the case of psychological measurements by adhering
strictly to the tenets of scientific research and measurement, treating the problem of
desirability is more challenging. On tests of intelligence, though the test items like an
arithmetical ability, logical numerical ability, and so on, lessen the scope of the desirability
of responses on part of the subjects, the subject(s) may resort to guessing, which may again
fail the purpose of psychological research and measurement.

6. Variability of Human Attributes Over Time


Various human attributes, like intelligence, personality, attitude, and so on, are likely to vary
over a period of time, and sometimes even hours are sufficient to provide scope for such
variations. Psychological attributes are highly dynamic and they continuously undergo
organisation and reorganisation. To capture these fluctuating attributes in terms of exact
numbers is really an uphill task for any researcher. Apart from this, this variability of
attributes acts as a threat to the validity of psychological research and measurement.

7. Problem of Quantification
It is questionable whether numbers are eligible and capable enough to denote all the
psychological attributes. Quantitative measurement has its own limitations because
everything that exists may not always exist in some amount, and even if it may, we may find
it difficult to assign this amount a number that exactly captures its meaning and essence.
For example, various emotions like disgust, jealousy, grief, surprise, and so on, are difficult
to quantify. A researcher may find it difficult to decide various levels for such variables in
quantifiable terms, and the choice of a number that is appropriate for a particular level of
such attributes may become even more difficult.

ERRORS OF MEASUREMENT
To error make mistakes is quite natural. Even in our day-to-day life, we commit a number of
errors; some of them because of our nature and others because of sudden situational
changes or chance factors. This implies that errors are as necessary as purity, because
nothing can be labelled as ‘complete’ or ‘perfect’ in itself. Though errors are unavoidable,
they can certainly be reduced to a minimum by taking precautions.
The same is true for measurement be it physical, mental or psychological though the errors
of measurement in mental science and psychology are much more prevalent than in physics
because the nature of the subject is comparatively more complex. For example, four
individuals A, B, C and D were asked to measure the length of a 50-inch-long table. The
length measured by A was 44 inches, and 47 inches and 48 inches by C and D, respectively.
Although the table is a physical object, the four persons did make error in measuring its
length. The error in physical measurement can be calculated by mathematical numbers.
In contrast, for mental sciences and psychology, both the measurement and errors of
measurement are difficult to assess because in addition to the personal characteristics of
the subject, the situational and chance factors also influence the results. The errors of
measurement are calculated by different statistical measures.
But what is meant by errors of measurement? Generally speaking, the difference between
the actual score of a person on a certain job and the obtained score by him is called error of
measurement.
For instance, on a certain test, the intellectual capacity of a child is calculated as 115 but his
actual intelligence quotient (IQ) is 120. This error is due to the unreliable test used and this
is called the error of measurement. To make it clearer, let us consider another example. In a
language test of English, a student scored 70 marks out of 100, but in another language test
of English, he scored 50 marks out of 100. This is attributed to error either in test or in
measurement.
Errors of Measurement refer to the difference in an individual’s hypothetical true score and
the actual score obtained by him. Since exact determination of an individual’s hypothetical
true score is not possible, all the measurements carry an element of error.
The point to be remembered is that there exists an inverse relationship between error of
measurement and the confidence one can lay in the measurement, that is, the reliability of
the measurement. The lesser the error of measurement, the more will be the reliability of
the test (or any other measurement technique used). Thus, the reliability of test/instrument
can be estimated by error of measurement. It can, therefore, be concluded that the errors
relating to measurement are those that are usually committed while collecting data.

SOURCES/TYPES OF ERROR
Psychologists have divided the errors of measurement in Seven types:
1. Accidental/Chance Error
2. Systematic/Biased Error
3. Interpretative Error
4. Variable Error
5. Personal Error
6. Constant Error
7. Statistical Error
1. Accidental/Chance Errors
These errors, as the nomenclature suggests, can occur any time; for example, sudden noise
during test administration or physical inconvenience or pain experienced by the subject
during the administration, and so on, are certain interfering factors that can produce errors
in the results. As these factors are situational, the amount of chance errors varies. Also the
direction in which they work varies, that is, such errors can sometimes increase the
resultant and sometimes decrease its value. When the scores of students increase or
decrease, it is the chance factors that are at work. An important quality of chance factors is
that with repeated measurements, these errors diminish naturally.
Such errors are also inherent in the test/instrument being used and these errors are called
test centred errors. It is not possible for any test to be totally culture-free because no test
can contain all the items applicable to all types of cultures. Any test administered in two
random samples from a population yields different results and this difference in results is
attributed to the test-centered errors. This type of error is test-centred because items
related to different populations cannot be put together in one test.
The second type of chance errors are subject-centred errors. All the factors related to the
subject like health, motivation, will, attention factor, working ability, and so on, do matter
during the administration of the test. Any disturbance in these factors can bring about
significant difference in
the results.
The third type of chance error is assessment error. Generally, sufficient numbers of
examiners are not available to assess the subject’s result, which causes inconvenience as
well as errors of measurement. In material assessment, such errors are less, but in the
assessment of detailed or essay type responses, these errors are more prominent because
the ‘situation factor’ has to be considered. Thus, the degree of such errors is more.

2. Systematic/Biased Errors
As these errors are in a specific direction, they can produce quite misleading results. These
errors are a result of polluted thoughts, personal biases, improper morals, and so on, and
these errors can never be removed.
Under such errors, we first talk about personal errors which arise from mistakes made by
the tester during the test administration or lack of alertness on his part; for example, errors
made in reading the thermometer, stop watch or the test manual. Second, such errors occur
when in a haste, the tester starts or stops the stop watch early or when the start or stop of
the stop watch is delayed due to some reason. This shows its effect directly on the
experiment or results, and this destroys the actual aim of the study (or the experiment).
These errors can be removed by comparing the results obtained by several
testers/experimenters.
The second type of biased errors arise from carelessness. These errors are more evident
when, for example, the experimenter instead of writing 0.1 writes 1 or instead of writing 0.1
writes 0. 01, or instead of +ve (positive) he writes –ve (negative). These errors can actually
alter the results. In order to avoid these errors the experimenter and the evaluator should
be very careful while recording and calculating the results.
The next type is unavoidable errors. Social scientists are generally interested in humans, an
human behaviour is influenced by so many internal and external factors that it is perhaps
impossible to tell the number of these factors. The most evident factors can be controlled
by the scientists but there are still too many factors left uncontrolled, and these factors
leave their impact by altering the behaviour in some way or the other. There are many
factors which, we might think, cannot influence the particular behaviour but still they do
leave some impact. These unchecked/uncontrolled factors produce errors called
unavoidable errors. These errors can be reduced by controlling more factors.

3. Interpretative Errors
Interpretative errors occur due to the wrong or erroneous interpretation of the
(test/experiment) results or scores. Sometimes these errors occur because of a set of wrong
assumptions on the part of the researcher, and sometimes these errors occur due to some
inherent problems related to the numerical measurement of the psychological properties.
There are two shortcomings in the measurement of behavioural quality. First, there is no
absolute zero point. Second, it is not certain that whether all the points on the scale are
equidistant. Because of these two reasons, in measurement, there is no value or meaning of
data and results, until the results are interpreted by other persons.
In these types of errors, the examiner is usually not able to understand as to in which
group’s context is an individual’s score to be interpreted. In such a situation, generally, the
examiner evaluates inappropriately and also interprets incorrectly. Such mistakes are known
as ‘interpretative errors’. These are committed due to the misunderstanding by one or the
other regarding:
1. Which kind of group an individual has been compared to and
2. In what way the comparison between the individual and the group has been expressed.
For
example, if in a test, the scores obtained by the students of high school are compared to the
scores obtained by the students of Intermediate or those of class VIII, the difference in the
interpretation of results will be significant. Similarly, the difference would be evident in the
hypothesis.
These types of errors are related to the standardisation of the test. By standardisation, we
mean that a test should be administered to well-defi ned groups and the scores obtained by
these groups should be recorded because these records set the standard for a particular
test. For any test, the standards can be prescribed for many groups and these standards can
be explained in different forms such as occupation, geographical area, sex, age, school
grade, and so on. The main objective of this distribution is to compare an individual’s score
to appropriate group norms. In order to control such errors, the experimenter should
particularly keep in mind the time factor during the standardisation process. In order to
interpret an individual’s score, it is essential to understand two such groups to which an
individual can be appropriately compared. The first group is one, of which the individual
himself is a part (member), and his actions and abilities are compared to those of its other
members. The other group is that which the individual aspires for. For example, if a college
student is interested to be a lecturer, then it will be important to compare his educational
qualifications and test scores to those possessed by the existing lecturers of college. The
final problem that is to be considered regarding the interpretative errors is that the
individual’s obtained scores are converted into derived scores like mental age, IQ, standard
scores, and so on, so that they could be compared to a normative group.

4. Variable Errors
The second type of errors in psychological measurement is known as variable errors. These
errors are consequences of impurities arising due to different reasons and situational
factors. For example, if the IQ of a person is measured during three different occasions by
different persons, the results will vary. These errors are evident in psychological as well as
physical measurements. Snapping of lead during administration or any kind of noise, pain
experienced by the subject during administration, and even the incorrect instructions by the
test conductor, can alter the results. All these are situational errors and these are called
variable errors because the degree of error varies from person to person and if one person
is tested twice on different occasions, the degree of error of measurement would be less.
Such type of errors in test can be estimated through the amount of confidence one can lay
on a test (the test reliability).

5. Personal Errors
The third type of errors in measurement is called personal errors. These errors are due to
subjective element of an individual. For example, if four persons sitting in a car are asked to
read the speedometer, the difference in readings will be quite evident. Similarly, two
persons evaluating and interpreting the same response do so quite differently because each
has a unique perspective. In the same way, two testers evaluating one subject express their
results differently and if the same subject takes the same test twice on different occasions,
it gives different results. Such errors are personal errors. Thus, personal errors should be
checked out (kept in mind) during the measurement of behavioural qualities. Efforts are
made to reduce personal errors during testing and measurement through objectivity.

6. Constant Errors
These are based on the fact that most of the behavioural qualities are measured indirectly.
It is not possible in psychological measurement, like in mental ability, to dissect the person’s
brain and tell that this much is his mental ability or intelligence! Instead, here, the
measurement is related to internal abilities and qualities. It is evident that the score of any
individual on a mental ability test depends on his ability to read as well. The difference in
these two qualities gives rise to constant error. Such errors are called constant errors
because the amount of such errors remains constant irrespective of the tester. Such errors
are related to the test validity.
Thus, in order to reduce these errors, it is most important to find out whether the test
actually measures what it is supposed to measure, that is, is the test valid? Also, the
description of errors of measurement remains incomplete unless statistical errors are
mentioned.

7. Statistical Errors
Generally, statistical errors are divided into two types:
1) Errors of Descriptive Statistics: Under these are included standard errors and probable
errors.
2) Inferential Errors: Under these are Type I and Type II errors.

1) Errors of Descriptive Statistics


There are two ways of calculating errors of descriptive statistics. The first is the standard
error and the second is the probable error. Standard errors get quite affected by extremes
because standard errors are calculated using standard deviations.
Thus, the standard deviations in such cases are impure. In such conditions, probable error is
calculated. Both standard error and probable error indicate the amount of error or expected
error in the data obtained. After calculating these errors, we can also know whether these
errors are due to chance factors or due to deviation in variables. If these are due to chance
factors, the degree would be less. But if these are due to differences in variables, the degree
of error would be high. The determination of error depends to a great extent on the nature
of research and planning.
Sometimes we agree to take 5 per cent chance errors and sometimes it becomes difficult for
us to take even 1 per cent chance errors.
a) Standard Errors
Whenever the mean of a sample is calculated, the obvious argument is whether this mean is
appropriate or not, that is, what may be the difference between the sample’s mean and the
mean of population from which the sample is drawn. The difference between the two
means can be expressed through standard error. Standard deviation of the distribution of
statistics is called the standard error of that statistics. Therefore, standard error is that value
which implies the amount of variation in the mean and the standard deviation of a sample,
and the mean and the standard deviation of the population from which the sample is
drawn. The standard error is important in order to study and understand the characteristics
of the sample. Through standard error, we can know how near the sample is to the
population concerned. The less the standard error of a sample, the more it represents the
population. A good sample will show low standard error.
b) Probable Errors
If the difference in results of two observations on different occasions is less or if the
observations of two persons on a problem are not very different, then it is not considered
true or reliable, but a small deviation in two observations make it reliable. Thus, when we
try to know the truth through measures of variance, the result is probable error.
2) Inferential Errors
The following are two types of the inferential errors:
a) Type I error
b) Type II error

a) Type I Error
This type of error is committed when a null hypothesis (H0) is rejected and, in fact, it is true.
For example, the H0 is that there is no significant difference between the IQ of boys and
girls. Suppose, the calculated t value of these groups is less than 1.96, then H0 is rejected
indicating that there is some difference between the two groups; but if there is no actual
difference, then there is some difference between the two groups. But if there is no actual
difference in population means and difference is observed on the basis t value, then Type I
error is committed. This error is represented by α (alpha). There exists an inverse
relationship between α and the significance level. If the significance level is high, α will be
less, and if the significance level is low, Type I error will be more. For example, if the
significance level for any hypothesis is kept at 0.01, then there would be more chances of
committing Type I error, and if the significance level is kept at 0.05, then the chances of
committing Type I error would be comparatively less.
b) Type II Error
Such an error is committed when H0 is accepted and, in fact, it is false. For example, H0 is
that there is no significant difference between the IQs of boys and girls. Suppose the t value
calculated for these groups is more than 2.58, then the H0 is accepted but actually there is a
difference between the two groups. Thus, Type II error is committed. This error is denoted
by β (beta). Type II error has direct relation with the significance level. Higher the
significance level, more are the chances β of being committed; for example, Type II error will
be more when the significance level is 0.5 as compared to when it is 0.1. This is important in
psychological experimentation.
Both the types of errors are related to each other. It we try to reduce Type I error, we
increase the chance of Type II error being committed and, similarly, if we wish to reduce
Type II error, we are taking a chance with Type I error.

CONCLUSION
The process of assigning values to represent the amounts and kinds of specified attributes,
to describe (usually) persons.
o We do not “measure people”
 We measure specific attributes of a person
Psychometrics (Psychological measurement) is the “centerpiece” of empirical psychological
research and practice.
All data result from some form of “measurement”
The better the measurement, the better the data, the better the conclusions of the
Psychological research or application

You might also like