NON PARAMETRIC TESTS

In correlation and regression we computed characteristics of the sample in order to make conclusions about the characteristics of the population. Procedures which do this are known as parametric tests. However often the data does not approximate the normal curve and nonparametric methods have been developed which are free of the parametric assumptions in statistical operations such as "t" and "F".

The appropriate tests depend upon the nature and type of the subject data: NOMINAL DATA Data that is descriptive. Eg hair colour, name. Cannot be added together.

ORDINAL DATA

Scores. Eg 6/10,1/5

INTERVA/RATIO DATA

Temperature, achievement scores. Can be added together.
The range of popular tests are as follows:

 NOMINAL ORDINAL INTERVA/RATIO NOMINAL Chi square <---------- "t" test or analysis of variance (ANOVA) ORDINAL ^ Spearman's Rank Order Correlation <--- ^T INTERVAL/ RATIO "t" test or analysis of variance (ANOVA) <------- ^ Pearson's Correlation (Correlation)

Therefore, nonparametric tests such as chi square are used where one or both of the sets of data is nominal. In this part we will cover 2 nonparametric tests; chi square and the sign test. See chi square

See sign test

2