PARAMETRIC AND NON-PARAMETRIC STATISTICAL TESTS

The parametric and non-parametric statistical tests are commonly employed in behavioural
researches. A parametric statistical test is one which specifies certain conditions about the
parameter of the population from which a sample is taken. Such statistical tests are
considered to be more powerful than non-parametric statistical tests and should be used if
their basic requirements or assumptions are met. These assumptions are based upon the
nature of the population distribution as well as upon the type of measurement scales used
in quantifying the data.

The assumptions may be enumerated as follows:

1. The observations must be independent. In other words, the selection of one case
must not be dependent upon the selection of any other case.

2. The observations must be drawn from a normal distributed population.

3. The samples drawn from a population must have equal variances and this condition
is more important if the size of the sample is particularly small. When the different
samples taken from the same population have equal or nearly equal variances, this
condition we known as homogeneity of variance. Statistically speaking, by
homogeneity of variance mean that there should not be a significant difference
among the variances of different samples.

4. The variables must be expressed in interval or ratio scales. Nominal measures (that
is, frequency counts) and ordinal measures (that is, rankings) do not qualify for a
parametric statistical test.

5. The variable under study should be continuous. The examples of a parametric test
are the z test, t-test and F test.

A non—parametric statistical test is one which does not specify any conditions about the
parameter of the population from which the sample is drawn. Since these statistical tests do
not make any specified and precise assumption about the form of the distribution of the
population, these are also known as distribution-free statistics. The non-parametric statistics
do not specify any rigid conditions like parametric statistical tests although certain
assumptions are associated with them. For a non-parametric statistical test, the variables
under study should be continuous and the observations should be independent. But these
assumptions are neither rigid nor so elaborate as we find in the case of a parametric
statistical.test. The examples of non-parametric tests are the chi-square test, the Mann—
Whitney U test, Kendall's tau, Kendall's coefficient of concordance, etc.

A non-parametric statistical test should be used only in the following cases:

1. The shape of the distribution of the population from which a sample is drawn is not
known to be a normal one.
2. The variables have been quantified on the basis of nominal measures (or frequency
counts).
 3. The variables have been quantified on the basis of ordinal measures (or ranking).

Because non-parametric statistical tests are based upon frequency counts or rankings rather
than on the measured values, they are less precise, and are less likely to reject a null
hypothesis when it is false. That is why, a non-parametric statistical test is used only when
the parametric assumptions cannot be met. Some statisticians, however, argue that non-
parametric statistical tests are more powerful and have more merits than parametric tests
because their validity is not based upon the assumptions about the population distribution.
They further argue that the parametric assumptions are often ignored by the researchers
and there are evidences in certain parametric statistical tests like the t test and F test that
violation of assumptions, particularly when a sample is large, does not affect the power of
the statistical tests. Not only this, for some population distributions non-parametric
statistical tests are superior in power to parametric statistical tests.

The advantages of non-parametric statistics are as follows:

1. Simplicity and facilitation in derivation: Most of the non-parametric statistics can be

derived by using simple computational formulas. This advantage not lie with most of
the parametric statistics, the derivation of which requires an advanced knowledge of
mathematics.

2. Wider scope of application: Since non-parametric statistics as compared to

parametric statistics are based upon fewer and less rigid and elaborate assumptions
regarding the form of population distribution, they can be easily applied to much
wider situations.

3. Speed of application: When the sample size is small, calculation of non-parametric

statistics is faster than parametric statistics.

4. Susceptibility to violation of assumptions: In case of non-parametric statistics

assumptions are fewer and less elaborate than the case of parametric statistics.
Therefore, assumptions of non-parametric statistics are less susceptible to violation.
Not only this, these violations are easier to check and can be readily and
economically taken care of with the non-parametric statistics.

5. Type of measurement required: Non-parametric statistics require measurement

based upon a nominal scale and ordinal scale, whereas parametric statistics require
measurement based upon the interval scale and/or ratio scale. As treatments
associated with either nominal scale or ordinal scale are easier than treatments
associated with either interval scale or ratio scale, the parametric statistics have a
better case for applicability than the non-parametric statistics.

6. Impact of sample size: When size is 10 or less than 10, non-parametric statistics are
easier, quicker and more efficient than the parametric statistics. If the assumptions
of parametric statistics are violated for such small cases, the result is likely to get
badly affected. Therefore, for this sample size, non-parametric statistics are always
superior to the parametric statistics. The reader should note that as the sample size
 increases, non-parametric statistics become time-consuming, laborious and less
efficient than the parametric statistics.

7. Statistical efficiency: Non-parametric tests are often more convenient than the
parametric tests. If the data is such that it meets all assumptions of non-parametric
statistics but not of parametric statistics then non-parametric statistics have
statistical efficiency equal to parametric statistics. If both parametric and non-
parametric statistics are applied to the data which fulfils all assumptions of
parametric tests, the distribution-free statistics become more efficient with a small
sample size but they become less and less efficient as sample size increases.

Disadvantages:

1. The non-parametric statistics have lower statistical efficiency than parametric

statistics when sample size is large, preferably above 30.
2. If all assumptions of parametric statistics are fulfilled, Siegel (1956), and Siegel &
Castellan (1988) consider the use of non-parametric statistics as simply 'wasteful of
data'.
3. It is also said that the probability tables for testing the significance of non-parametric
statistics are widely scattered in different publications which, for a behavioural
scientist, is difficult to locate and interpret.

PARAMETRIC STATISTICS
Parametric and Nonparametric Statistics
A parameter, as we learned in an earlier chapter, is a population value. If all the scores of a
defined population are available and a mean is calculated, this mean is a parameter.
Similarly, the valiance and the standard deviation of a population are parameters. It may not
be possible to calculate population measures. They are still referred to as parameters. A
statistic, on the other hand, is a measure calculated from a sample. Whenever statistical
tests, parametric or nonparametric, are used, certain assumptions are made.

Nonparametric statistical tests are hemmed in by fewer and less stringent assumptions than
parametric tests. They are particularly free of assumptions about the characteristics or the
form of the distributions of the populations of research samples. Thus they are also called
distribution-free tests. As Siegel puts it, "A nonparametric statistical test is a test whose
model does not specify conditions about the parameters of the population from which the
sample was drawn.

Assumptions of Normality
The most famous assumption behind the use of many parametric statistics is the
assumption of normality. It is assumed in using the t and F tests, for example, that the
samples with which we work have been drawn from populations that are normally
distributed. It is said that, if the populations from which samples are drawn are not normal,
then statistical tests that depends on the normality assumption are vitiated. As a result, the
conclusions drawn from sampled observations and their statistics will be in question. When
in doubt about the normality of a population, or when one knows that the population not
normal, one should use a nonparametric test that does not make the normality assumption,
it is said. Some teachers urge students of education and psychology to use only
nonparametric tests on the questionable ground that most educational and psychological
populations are not normal. The issue is not this simple.

Homogeneity of Variance
The next most important assumption is known as the homogeneity of variance assumption.
It is assumed, in analysis of variance, that the variances within the groups are statistically
the same. That is, variances are assumed to be homogeneous from group to group, within
the bounds of random variation. If this is not true, the F test is vitiated. There is good reason
for this statement. We saw earlier that the within-groups variance was an average of the
variances within the two, three, or more groups of measures. If the Variances differ widely,
then such averaging is questionable. The effect of widely differing variances is to inflate the
within-groups variance. Consequently an F test may be not significant when in reality there
are significant differences between the means.

To return to the evidence on normality and homogeneity, Lindquist says, the F distribution is
amazingly insensitive to the form of the distribution of criterion measures in the parent
population. Lindquist also says, on the basis Of Norton's data, that unless variances are so
heterogeneous as to be readily apparent, that is, relatively large differences exist, the effect
on the F test will probably be negligible. Boneau confirms this. He says that in a large
number research the probability statements resulting from the use oft and F tests, even
when these two assumptions are violated, will be highly accurate. In brief, in most cases in
education and psychology, it is probably safer—and usually more effective — to use
parametric tests rather than nonparametric tests. It was concluded that parametric
procedures are the standard tools of psychological statistics, although nonparametric
procedures are useful minor techniques.

Continuity and Equal Interval & of Measures.

A third assumption is that the measures to be analyzed are continuous measures with
equal intervals. This assumption is behind the arithmetic operations of adding, subtracting,
multiplying, and dividing. Parametric tests like the F and t tests of course depend on this
assumption, but many nonparametric tests do not. A rank-order method, for example, may
take no account of the continuity and equal intervals of measures.
Despite the conclusions of Lindquist, Boneau, Anderson, and others, it is well to bear these
assumptions in mind. It is not wise to use statistical procedures-or, for that matter, any kind
of research procedures — without due respect for the assumptions behind the procedures.
If they are too seriously violated, the conclusions drawn from research data may be in error.
To the reader who has been alarmed by some statistics books the best advice probably is:
Use parametric statistics as well as the analysis of variance, routinely, but keep a sharp eye
on data for gross departures from normality, homogeneity of variance, end equality of
intervals. Be aware of measurement problems and their relation to statistical tests, and be
familiar with the basic nonparametric statistics so that they can be used when necessary.
Also bear in mind that nonparametric tests are often quick and easy to use and are excellent
for preliminary, if not always definitive.