Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

Longitudinal Vs.

Cross-sectional Analysis
Dr. D. R. Weerasekera
Department of Statistics, University of Colombo
The survey literature distinguishes cross-sectional and longitudinal designs, but
most surveys conducted in practice are based on cross-sectional designs. The
limitations of these cross sectional surveys are that they are confined to a specific
point in time. That is, they provide us with a snapshot of a sample of a population
at a single point in time. Since population characteristics constantly change over
time, cross-sectional surveys for such situations do not reflect the actual situation.
This raises the issue of measuring social characteristics over time in a single study
the objective of a longitudinal study. The study of change over time is called
longitudinal analysis. In this sense, longitudinal studies involve the study of a
process of change over a period of time. Such trends can also be observed by
comparing the results of surveys that are conducted in separate years. The aim of
this paper is to point out the importance of conducting longitudinal surveys for a
variety of situations, and the importance of applying the correct statistical
techniques for data obtained from longitudinal surveys.
The simplest type of longitudinal analysis of survey data is called trend analysis,
which examines overall change over time. Trend analysis has some significant
limitations. While it can reveal changes, it gives us little insight as to how or why
the changes have taken place. One possibility is that individuals change their
attitudes or behaviors as they move through the life cycle. Another possibility is
that people change because of new circumstances. A major crisis, such as a war or
depression may result in such changes across all age groups. Changes in technology
may have a similar impact.
Cross-sectional surveys often provide data that reveal little change from one year
to the next. This can be seen especially, when we consider figures on
unemployment or characteristics of poor households receiving Samurdhi-benefits
implying that the same households remain poor over time. On the other hand,
several longitudinal studies point to considerable variation over time among poor
households and reveal that they are in a transition state. Thus, some households
typically move between states of poverty; others move from a state of poverty to
relative affluence; others slide into states of poverty. Longitudinal surveys further
enable us to detect and monitor variations and trends among individuals, as in the
case of variations in salary among workers. Here the value of collecting data at
several points in time cannot be measured as changes in the job patterns and
incomes of people can be monitored effectively only at an individual level.

Repeat surveys, on the other hand, offer a distinct advantage as they enable us to
capture the net effect changes. In the case of an opinion poll, such net effects
might be expressed as overall increase or decrease in the number of people who
report an intension to vote for a specific party. By repeating the survey at a
different time and asking fairly similar questions, it enables us to collect
information that can easily be compared. However, repeat surveys collect data
from different respondents and thus we are unable to determine the gross changes
in the intention to vote. This limitation is clear from the inability of cross-sectional
and repeat surveys on voting patterns to provide detailed information about
respondents who are undecided about whether they vote in an election.
One example of a cohort study in a country is that a birth cohort is being followed
over a ten-year period in order to monitor changes in their lives. In such a survey,
the same respondents are interviewed at different times during the study. Here, (a)
data are collected at two or more different points in time; (b) the same sample is
interviewed at distinct points in time; and (c) data from the respondents are
compared across these time points in order to monitor patterns of change and
promote social understanding.
In panel studies, the same people are interviewed at two or more points in time.
Since the sample is the same, any changes we observe are not a result of sampling
error. Panel studies, however, have problems of their own. For one thing, they are
generally very expensive, as it is not easy to keep track of respondents. Further,
despite our best efforts, we may not be successful in all of our attempts to recontact respondents, especially if the study is conducted over a long period of time.
Those who drop out of the panel (by moving, dying, refusing to continue, etc.)
may have changed their attitudes and behaviors from those who remain.
The main disadvantages of longitudinal surveys are that they are costly.
In longitudinal studies, individuals are measured repeatedly through time or the
same variable is measured repeatedly. Therefore, longitudinal data require
sophisticated statistical techniques because the repeated observations are usually
(positively) correlated. Correlation must be accounted for to obtain valid
inferences. Sequential nature of the measures implies that certain types of
correlation structures are likely to arise. One approach to analyzing repeated
measures data is to consider extensions of the one-way ANOVA model that
account for the covariance. That is, rather than assuming that repeated
observations of the same subject are independent, it allows the repeated
measurements to have an unknown covariance structure. To do this, we can easily
use the SAS procedure, PROC MIXED, an extension of PROC GLM which allows
clusters of correlated observations.

You might also like