Rstanarm output


#1

Hi, I have a couple concerns.

First, rstanarm (and rstan in general) spews out a ton of output, every 10% of iterations in each chain, the elapsed time, just lots of crap which as a user I’d rather not see. Maybe if it’s gonna take more than a few seconds it would be good to have one or two lines of output so that the user won’t freak out. But all these screenfuls, that too much.

Second, I wonder if we could change the header of print output to be more similar to what lm/glm does.

Here’s an example.

From lm:
lm(formula = post_test ~ treatment, data=electric, subset=(grade==4))

From stan_glm:
family: gaussian [identity]
formula: post_test ~ treatment

I prefer the way lm does it, because sometimes we’re just looking at the print output and it’s very helpful to see the name of the data frame and any subsetting.


#2

It is refresh = BIG_NUMBER in rstanarm <= 2.17.x and refresh = 0 in rstan{arm} >= 2.18.x to curtail the progress output. My PR to do time-based updating is still in limbo.

To print out the relevant part(s) of the call in one line, we have to decide what is relevant.


#3

Also there’s a problem when we ask rstanarm to print out the priors it was using. It doesn’t give the full priors on the original data scale.