mombf package vignette

This manual shows how to use the mombf library to compute Moment and inverse Moment Bayes factors (Mom BF and iMom BF, respectively). The intuitive appeal of Mom and iMom BF is that they represent prior beliefs under the alternative hypothesis which are fundamentally different from those under the null hypothesis. Mathematically, when the null hypothesis is true they present better convergence rates than BF resulting from most standard procedures. When the alternative hypothesis is true, they present the same convergence rates as most standard procedures. The routines compute exact BF for linear regression models, and approximate BF for generalized linear models. Approximate BF can also be obtained in other situations where the regression coefficients are asymptotically normally distributed and sufficient. The library also contains routines to evaluate the prior density and to elicit the prior parameters by specifying the mode a priori of the standardized regression coefficients. In Section 1 we briefly review the definition of the Mom and iMom priors, and we present routines to evaluate them. In Section 2 we analyze Hald’s data with linear models and compute Bayes factors to assess whether some predictors can be dropped from the model. Section 3 shows the analysis of some simulated logistic regression data.