CERG Home Page  

CERG Resources

Particular when doing longitudinal or follow-up studies, you will measure a group of subjects more than once.

The basic research question is:

Can the differences between the two measurement times be attributed to chance or are the differences statistically significant?

Your null hypothesis is that there is no differences between measurement times

If your data is ratio or interval then you have:
For 2 repeats - the t-test for correlated groups

For more than 2 repeats - Analysis of Variance (ANOVA)

If your data is interval then you have:
For 2 repeats - the Wilcoxon test

For more than 2 repeats - Friedman's ANOVA

If your data is nominal or categorical then you have a problem
You might be able to use a Chi-square test but this particular problem is not what is is defined to do.

By stretching things a little, you could use it, but you might find the results will be rejected if you submit a paper to a professional journal.