Trouble to interpret rmANCOVAs
Hi all,
My question concerns a study that was performed to assess fear extinction learning, i. e., the "unlearning" of previously acquired fear. The study has different phases: habituation, acquisition (basically classical fear conditioning with 3 different conditioned stimuli), extinction, retrieval, and reinstatement test. Skin conductance responses were reported as fear measure. I think it is not necessary to understand what the different phases are exactly to understand my question, though.
I've conducted rmANOVAs for each phase to assess for fear learning and fear extinction. For retrieval, the analysis revealed significant effects of trial, stimulus, as well as a significant trial * stimulus interaction. For reinstatement, there were no significant effects. So far so good.
I am, however, furthermore interested in the effects of different sleep variables (such as time spent in REM or deep sleep) on fear extinction learning, especially concerning the retrieval and reinstatement phase. Thus, I've additionally conducted rmANCOVAs for retrieval and reinstatement test with different sleep variables as covariates. For retrieval, all significances disappeared, while for reinstatement, there is now a significant Stimulus * Context * Trial interaction (the stimuli were in part presented in different contexts during the experiment).
I now have trouble to interpret the results of the ANCOVAs. You've already referred me to the between subjects effects table but I don't really understand it. It contains no significant p-values, but there must be some effects of the sleep variables due to the differences between the ANOVAs and the ANCOVAs, or am I wrong? If you can think of any post hoc analyses I can/need to conduct now in JASP, I'd be glad if you could let me know.
Also: I know that there are probably more suitable analyses to assess my research questions, but this is the one I agreed upon with my supervisor and I don't have much time left, so I kinda have to make the best out of it.
Thank you very very much in advance! If you need any further information, please let me now.
Comments
Hi Dexterama. I think you should not expect the ANOVA results to be compatible with the ANCOVA results, since the two types of analyses ask and answer different question. Specifically, each ANCOVA draws the following type of conclusion. "Had I held-constant the values of my covariates, it's estimated that my means and variances, and the accompanying effects of the independent variables on the dependent variables, would be NOT as indicated by the ANOVA results, but would be as follows: . . ."
R