Bayes factors - corresponding method to OLS regression with individually clustered SEs

[Nepo Hansen]

Dear all,

I'm analyzing experimental data for a thesis project where individuals in an experiment make several choices. I.e., I have repeated observations (3 obs. per i = 1,...,N with N = 150) for each individual in the dataset. I used OLS regressions with standard errors clustered at the individual level to analyze the data. However, reviewers asked me to provide Bayes Factors.

What would be the corresponding method to be used with the bayes: prefix in Stata to the OLS with clustering at the individual level? I assume that the Bayes Factors I receive when ignoring the repeated measures nature and just use "bayes : reg" are not correct?

I'm not really acquainted with Bayesian statistics, please excuse this maybe stupid question.

Best;
N. Hansen

[Rich Goldstein]

I am not a Bayesian (just to get that out of the way) but there are simple ways to approximate a Bayes Factor from a p-value; for a quick intro to this, see https://www.bayesianspectacles.org/f...-3p%E2%88%9An/

[Nepo Hansen]

Hi Rich, Thank you very much for your reply! This will be what I do alternatively then. However, I'd prefer to get it done with Stata's Bayes functionalities if there is some way.

[Leonardo Guizzetti]

It seems to me that Bayes factors are a reasonable request only if you conducted a Bayesian analysis up front. You may consider pushing back against the reviewer's request for this.

[Carlo Lazzaro]

Nepo:
https://projecteuclid.org/journals/a...0-AOAS362.full might be interesting (ETA: I'm not familiar with this stuff; just found it [open access] out from the web).

[Nepo Hansen]

Dear Leonardo and Carlo, Thank you for your replies! I think, it might be more appropriate to think about equivalence testing then rather than going to the Bayesian world. :-)