bayes factor hypothesis test in ANCOVA as a alternative for linear mixed model
Hi, everyone
I want to test whether or not two groups of people have differences in one continuous outcome. But those data are collapsed from 3 different datasets, and I also have 2 categorical variables and 2 continuous variables that need to "control".
I first tried linear mixed model, one continuous outcome as DV, group and 4 covariates as fixed factor, and dataset label as a random factor(both for intercept and group slope). I find a nonsignificant p value.
Then I want to try the Bayes factor hypothesis test for the evidence of the null hypothesis. I noticed there has no Bayes factor test implemented in "Bayesian linear mixed model" in JASP and some posts recommend the "ANCOVA" module. So I was wondering:
(1) Could I do Bayesian ANCOVA to test the group difference? If I'm in the right way, should I enter the categorical covariate in fixed factors and continuous covariate in covariate box (this box has a label indicating only continuous measure....) then adding all of them to null model ? Alternatively, I can enter all covariates in covariate box...
(2) Some classmates also suggest me to decomposite the random factor(dataset label) into several dummy variables (ie, dataset1 = 1, others = 0; dateset2 = 1, others = 0; and so on) then treat them as covariates in Baysian ANCOVA analysis. Did this sound more reasonable?
Any help is welcome, thanks a lot for reading!
Comments
Hi @carl559 ,
Since you initially used mixed models, this implies that you have nested observations within some categorical variable (e.g., within one participant, or experimental group - if I understand correctly this is "study label" for you). If you want to run a comparable analysis in the Bayesian ANOVA module, I would recommend doing so in the Repeated Measures analysis. However, your data would need to be in a different format: for the mixed model module, your data needs to be in the long format (so I'm guessing this is how you have your data), but for the RM ANOVA, your data needs to be in the wide format. If you do not know how to go from one format to the other (JASP cannot do this yet, unfortunately), I can help you with this (and make a suggestion for how to set up your JASP analysis. To do so, I would need the data set and more description about the data structure (what does each variable/column represent, what does the nested structure look like). If you want, you can send the data by email to johnnydoorn <at> gmail <dot> com.
Kind regards
Johnny