Bayesian Logistic Regression on JASP: documentation, choice of prior

PM_Stat

Hello,

I'm using JASP 0.18.2. In this version, it is possible to conduct Bayesian logistic regression. But I don't find documentation about this analysis on JASP. I only found the paper of van den Bergh et al. (2021) about multi-model linear regression.

I suppose that the principle is similar, but I see that the priors are different. Especially, JZS prior is not available for Bayesian logistic regression. JASP automatically ticks the CCH prior (alpha = .5, beta = 2, s = 0). Does this prior corresponds to the default non-informative prior ? If yes, do you have documentation allowing to justify such a choice? If no, which default prior do you recommend ?

Thanks

Best Regards

vandenman

Hi,

The prior we use in logistic regression is based Li & Clyde (2018). There is not really a notion of non-informativeness for the priors in logistic regression. Instead, other properties are examined (e.g., model selection consistency), see section 4 of the referenced paper.

Hope that helps!

Don

Li, Y., & Clyde, M. A. (2018). Mixtures of g-priors in generalized linear models. Journal of the American Statistical Association, 113(524), 1828-1845. https://doi.org/10.1080/01621459.2018.1469992