I am struggling to find the most effective way to check randomization within my experiment. I have a single factor design with two moderators and two DVs. My IV is type of persuasion (0 direct persuasion, 1 self persuasion), the dependent variables are purchase intention and perceived persuasive intent (both measured as scales). The moderating variables are knowledge (recoded to low/high) and involvement (recoded to low/high). I measured age (scale) and gender (f/m) as demographics. The only manipulation was type of persuasion, and the participants were equally and randomly divided into that variables' conditions.
Could anyone help me figure out a way to see if randomization was successful i.e. if I should use age and gender as covariates in the MANOVA I am about to run?
I'm new to statistics, but my supervisor couldn't figure it out either and said just test randomization for your main manipulation type of persuasion by doing a chi square test. I'm not sure whether to do these one by one though, or put age and gender on the same axis, and what results to report (Fishers Exact, Cramer's V and such).
Thanks a million!
Best Answer
I assume you randomly assigned participants to the two experimental conditions (direct or self persuasion). Assessing how well the randomization worked would presumably involve testing for association between that grouping variable and each of your other variables you want to use as predictors. These would I assume be done one at a time. In cases where you're using binary predictors, a Fisher exact test could be used, though you might want to look into the implications of dichotomizing predictors measured on scales, as that's not necessarily the best thing to do. For "continuous" predictors, a standard t-test or correlation would yield the same results in terms of significance or p-values.