Bayesian ANCOVA in JASP
I have a question about Bayesian ANOVA (ANCOVA). I have the following data.
Dependent is the dependent variable, Within_sub is the within-subjects factor, and there are 3 levels (1.54, 2.45, 3.85). Each subject (subject: 1 ~ 10) participates in all levels, and there are 60 trials per level per subject. Here, in a typical Repeated Measures ANOVA in our field, the within-level, between-trials mean of the subject is taken as the representative value for each subject, and the between-subjects mean is compared across levels.
However, I would prefer to analyze the data from all trials, instead of using the subject's between-trials mean as the representative value for each subject. This is because I know that Covariate is correlated with the dependent variable, and I want to know not only the effect of the factor on the dependent variable but also the effect of Covariate on the dependent variable. So, instead of Repeated Measures ANOVA, I used Bayesian ANCOVA, using Subject as Random Factors, and I got the following results.
The questions are
Is it correct to use Subject as Random Factors?
Is this included in the linear mixed-effects model in a broad sense?
What is the difference between this and Repeated Measures ANOVA?
Why is it not possible to specify Random Factors in ANCOVA for the frequentist version in JASP?
If the structure of JASP data follows the rule of "1 row, 1 subject/unit", I am worried that this analysis may be wrong.
Please excuse my poor question.
Thank you in advance.
Comments
Hi RadHead5,
Hmm I will ask some other team members for their input. We do have linear mixed model functionality in JASP, but that does not produce Bayes factors at the moment. However, the Bayesian AN(C)OVA as implemented through the BayesFactor package is really a mixed model. So you could look at some of the documentation for the BayesFactor package to see how it's done there. To be continued...
E.J.