Bayesian Repeated Measures ANOVA with Many Models

RBJ

Hello all,

I'm trying to analyse some new experimental data with a Bayesian repeated measures ANOVA. I have 2 repeated measures factors (with 2 levels each), and 3 between subjects factors (also categorical with 2 levels each). Everything works fine if I use my 2 repeated measures factors and only 2 of the between subjects factors, but if I try to include all 3 between subjects factors the analysis takes far too long to run. It seems like it will never finish running on my laptop. Is there a workaround for this under the Bayesian framework? Or am I stuck doing a frequentest analysis?

Many thanks for any help you can provide,

RBJ

JohnnyB

Hi RBJ,

You can try tweaking the number of samples used for MCMC. This option is under the 'advanced' menu. I would suggest using the default setting when you actually want to report your results though, as decreasing the number of samples typically decreases the numerical stability of the result.

You can also try to run it on a different computer - how many participants do you have? Because a 2x3 mixed ANOVA should not be that disastrous (at first I thought you had something really funky like 4x3x3x3x3 ;)).

Kind regards

Johnny

RBJ

Thanks Johnny,

I already tried dialing the MCMC samples back to 100 and it was still infeasibly slow. It's also running on pretty good hardware.

As I understand the study, I don't think that it would be 2x3. I believe that with 2 repeated factors and 3 between subjects factors all with 2 levels this would be 2x2x2x2x2. Am I thinking about this the wrong way? Does that change your recommendations?

Best regards,

RBJ