Default prior width for Bayesian ANOVAs in JASP
Hi all,
I have a question concerning the default Cauchy prior width used for Bayesian ANOVAs in JASP. For t-tests, a prior width of 0.707 is recommended by e.g. Rouder et al., and this corresponds to the default width implemented in JASP. For ANOVAs however, the default width is set to 0.5. I've been having trouble researching what the reasoning behind it is. Does anyone know/can anyone point me to a corresponding paper? Any help would be greatly appreciated.
Comments
It is set so that it gives the same result as the default t-test in case of a one-way ANOVA with two levels. The reason that the 0.707 is not the same as the 0.5 is because the ANOVA is parameterized slightly differently. Yes, this is very confusing. Maybe it is mentioned in the Rouder & Morey papers on ANOVA (they have several)
E.J.
Thank you for your quick reply. I'll have a closer look at these papers.