Hi
I have four groups of variables A, B, C and D. The variables A and B also have (natural) sub-groups (say A1, A2, A3 and B1, B2 and B3). I want to compare them as follows:
A - B - C - D (using ANOVA for 4 groups)
A1 - B1 (using t test for two samples)
A2 - B2 (using t test for two samples)
A3 - B3 (using t test for two samples)
The data overlaps (because of the sub-groups). Is it (statistically) ok to do such comparisons?
Or should I remove step one and go with something like this:
C - D (using t test for two samples)
A1 - B1 (using t test for two samples)
A2 - B2 (using t test for two samples)
A3 - B3 (using t test for two samples)
Note: This is not experimental data. Raw frequencies are standardized (z-scores), the data is related to language studies (corpus linguistics to be specific).
Thanks for your valuable comments.