
Message/Author 


Dear Drs. Muthen and Muthen, I am currently trying to run a multigroups (2 groups) path analysis, where the predictor and outcome variables are both continuous, and the mediator is a three category ordinal variable. I was wondering how best to input my mediator, in particular whether I should list this in the input as a categorical variable? I understand that if a categorical variable has less than 5 categories it should not be treated as continuous, but I wondered whether that is the same if treating the variable as a mediator? I have your book, and have been reading the section on binary and ordinal variables. I don’t think the input provided in 8.26 is appropriate in my situation as the outcome variable in the example is categorical and mine is continuous, is that correct? I’d very much appreciate any guidance, Best wishes, Nathalie 


You can use the Table 8.26 input with the modification that you don't put your outcome on the Categorical= list. You can also try the other alternatives discussed in Table 8.29 for which inputs are given on our book example web page http://www.statmodel.com/mplusbook/chapter8.shtml If the 3category ordinal mediator does not have a large percentage in the lowest or highest categories the difference between this and treating it as continuous is probably small. 


Many thanks, that is very helpful. If I can just clarify one point, my mediator does indeed have a larger proportion of participants in the lowest category. Are the methods in table 8.29 more appropriate in this situation? Many thanks again, Nathalie 


I would use WLSMV or Bayes because the latent mediator approach (as opposed to observed mediator) might be more appropriate with such a mediator distribution. 

Back to top 

