Operating System: windows 7
Interface Version:rstanarm (Version 2.15.3, packaged: 2017-04-29 06:18:44 UTC)
I had a model from which I meant to draw 500 samples
Speed_rep_D <- posterior_predict(gamma_all,newdataD,draws=500)
but happened to get much more - couldn’t find that explained in the help fpr posterior_predict.
By the way, here I had one factor with three levels and two random components and really were interested in comparing the posterior_predicts corresponding to the factor levels - In WinBUGS I might use “step” to see the difference directly, here I get predicted draws corresponding to the factor levels and compare outside of RStan - that cannot be the right thing to do?