I have been conducting all of my analyses on one DV (Quality) using a repeated measures ANOVA, with two variable (Time and Gender) at two levels each. However, it occurred to me today, that this might be incorrect, and here's the problem:
Though Gender is one of my variables, the truth is that participants don't find out gender until Time two. The setup is such that they read two essays (A and B) and evaluate them both (Quality). Then they find out that each of these essays is Male (M) or Female (F) and rate them again for Quality, however, I had counterbalanced the essays so that sometimes they would see AM or BM (or AF / BF).
Thus it seems that using a normal repeated measures ANOVA, the pairing would be wrong so the model assumes that A always goes with F and B always goes with M, so it's taking the difference between A and F and analyzing it agains the diff between B and M, correct?
MY ROUGH SOLUTIONS: Since it's nearly impossible for me to tell which essay was paired with which gender, my solution was to take the average of essay A and B and make a new variable (AB), and compare F and M to AB. However, it doesn't seem like I can do that using the normal model (at least not in SPSS).
Manually, I can do separate paired t-tests, or I could subtract the Grand Mean of (AB) from both F and M and then do a paired T-test on those, but again, that doesn't seem right.
Unless, maybe it is valid?
Or, would I just put in Time as a 3 level variable, and look at the pairwise contrasts after do the repeated measures without crossing it with Gender?
VALID SOLUTION? Is there any way to do a repeated measures ANOVA with all of these variables without having to do manual calculations or series of t-tests in which I would be using the wrong dfs?
PS: I use SPSS