Univariate GLM Model - asking for ETASQ PARAMETERS change effects?

[ComPsy]

Hi there,
I have a question regarding univariate GLM models. I have entered in three factors and as a covariates and asked for estimated means comparisons. I wasn't interested in interactions at first, I just wanted to see (a) one of my factors predicted the DV and (b) if my estimated means were different for those factors. I ran all those models a while ago (I have 15 models), and I was just redoing them and asking for the ETA squared as I realized I need to report it, but asking for this changes what I am getting in my main corrected model in terms of main effects.

Well, I reran this and found my results change when I ask for the ETA squared The factor that was initially significant in the corrected model section no longer was and other factors were. Can anyone explain why? Did I go from simply comparing estimated means to a full regression model? I am just not sure how asking for those would change my entire corrected model.

Can someone clarify this issue?
Thanks!
A

 

[mathgeek10]

This should not be the case. As long as your IVs and DVs are the same and your model is set up the same, your effect shouldn't change. Make sure you haven't made any changes to the data files. Also click "paste" and look at the syntax code from each of the models to make sure that the only thing you changed was asking for the eta squared info.

Also, I'm not really sure what your model intails, but it is genearlly always a good idea to look at the interaction. If one exists a long with a main effect, it will substaintially change how you interpet your results.

How this helps some!
Good luck