SPSS Post-hoc tests for Mixed Design ANOVA - eek!?

#1
Hey everyone! This is my first post, so be kind!

I was hoping someone may be able to help me. I've conducted a 2x3 Mixed Design ANOVA in SPSS, and have gained a significant effect for one of the variables. I'm trying to find out where the significance lie using post hoc comparisons. However, I'm not really sure what test to use! SPSS has worked out a Bonferroni comparison, which shows that only one of the comparisons is significant. However, a pairwise comparison using T Tests shows that 2 pairs are significantly different. Anyone have any idea which to believe? If it makes any difference, it's a within subjects variable.

Thanks for reading this!
 
#2
There are three main tests that people use for mulitple comparisons, Bonferoni, Tukey, and Fisher. Each of these tests has a different procedure to adjust the error rates of individual comparisons (group A vs group B) so that the error rate of all of the tests is at the level you want (often people do tests at the 95% level). If you do not use these adjustments, your errors can "gang" up on you.

For instance, if you are doing 6 tests, group A vs B, group A vs C, etc. and all the tests are done a the 95% level then the probability that all of these tests is correct is .95^6=.73. So when you thought you were doing a test with significance level 95% you are actually doing a test with a significance level of 73%.

Thus it is very important to use the multiple comparison procedures. Bonferoni is the most conservative of the tests whereas Fisher is most aggressive (generally) and Tukey is in the middle. Any one of these tests are fine to use as long as you say which one you have used.

Dave