There are various statistical techniques for analysing data. To choose an appropriate technique of statistical analysis in the challenging task to a research worker. It has two main functions:
1. Interpretation of results, and
2. Presentation of data.
The major types of tests are employed for analysing data so as to interpret the results. There are:
(A) Parametric statistics or tests, and
(B) Non-parametric statistics or tests.
A researcher has to select either of these approaches for analysing his own research data. The
following are the criteria for choosing an appropriate statistical approach.
(A) Considerations for Parametric Statistics
This type of statistical analysis may be employed effectively in the following conditions:
1. Probability or representative sample has been employed in the investigation.
2. Variables of the study can be qualified at interval scale.
3. Specific assumptions are fulfilled. The obtained data are normally distributed or not free distribution.
4. The population of the study has been clearly defined.
5. Objectives of the research study.
Under this approach the following statistical techniques are employed :
(a) To study the descriptive relationship of two or more variables:
a1 – Pearson’s product moment method of correlation (two variables)
a2 – Multiple correlation (more than two variables)
a3 – Partial correlation (more than two variables)
a4 – Factor analysis-extracting factors or estimating psychological or factorial validity of tests.
(b) To analyse the functional relationship of the variables:
b1 – Main effect of two treatments ‘t’ test
b2 – Main effect of more than two treatments F-test
b3 – Interaction effect of two or more variables-Two or more ways analysis variance techniques
b4 – Gain or loss of more than two treatments-Analysis of covariance and correlated ‘t’ test.
(B) Considerations for Non-parameteric Statistics
This type of statistical analysis may be used effectively in the following situations:
1. When non-probability sample is selected in the research study.
2. The variables of the study are quantified at any level of measurement, mainly, nominal and ordinal scale. It may be in the discrete form.
3. No assumption is required for this approach.
4. Free distribution of data, may be skewed or may be normally distributed.
5. Objectives of the study.
In this approach the following statistical techniques are generally used :
(a) To study the relationship of two or more variables:
a1 – Spearman Rho correlation in small sample not in large sample for two variables. Data are available at ordinal or internal scale.
a2 x2 and contingency correlation. It is used when two or more variables are taken. The data may be nominal or ordinal scale or interval scale.
a3 – Analysis variance.
(b) To analyse the difference between two or more groups:
b1 – Median test for small test.
b2 x2 test for large sample also for small sample.
b3 – Run test and U-test when data are on ordinal scale.
b4 – Sign test.