I am new to this forum, so I hope I am posting in the right place.
I have a question I hope someone can help me with.
I hope to examine differences between IV "i1" and DVs "d1", "d2", "d3", and "d4". I hope to use a MANOVA to do this. However before I do this main analysis to help shed light on my research question I am mindful that there are a number of potential covariates I should control for, let's call these "c1", "c2, "c3", and "c4".
To confirm whether I need to run a MANCOVA (controlling for these potential confounds) instead of a MANOVA I want to conduct some preliminary analysis on the data. Specifically, I want to know whether these potential confounds are actually present in my data set. Previous studies have done similar preliminary work, and have conducted a number of MANOVAs to determine whether these confounds varied across our "i1"levels (let's say there's 2). To determine this our original IV("i1" ) from the main analysis become the DV and the "c1" is the IV. A separate MANOVA is conducted with "i1" levels serving as the DVs and "c2" is the IV. And another ANOVA is conducted with "i1" levels serving as the DVs and "c3" is the IV, and so on. Only the confounds where a significant difference is found between groups (say "c1" and "c3" but not "c2" or "c4") is then controlled for in the main analysis' MANCOVA.
My question/concern is when conducting so many separate MANOVAs in the preliminary analysis phase would I have to adjust the p value (e.g., make a Bonferroni adjustment) at the multivariate level for each MANOVA to account the number of tests I am proposing to run in this preliminary stage of the analysis process? Previous literature adopting this approach have not made such an adjustment. If I don't have to I guess I would like to clarify my understanding as to why this is the case.
I hope I've been able to explain that in a way that is understandable.
Any help or advice would be most appreciated.
Advertise on Talk Stats