search

danlwarren / RWTY

R We There Yet?
30 stars 17 forks source link

add a global ess(chains) function #141

Open roblanf opened 4 years ago

roblanf commented 4 years ago

most people would like this. Just returns a table with all chains and all parameters (incl. tree topology) and their respective ESS values.

Also useful for the R markdown file. We could even just include the pseudo-ESS (just calculated with coda on the topo.dist.mcc column) as well as the approximate ESS for trees.

And we could consider a column with warnings for ESS values <200.

roblanf commented 4 years ago

note: right now @danlwarren has something that does this in the anlayze.rwty() function. All that's really required is to take it out of there and make it a user-facing function with documentation.

roblanf commented 4 years ago

Also - we could add the Beiko et al burnin method to this. So ESS's are calculated from global burnin (so everything's comparable), but we could have a column that indicates according to the Beiko method when each parameter started sampling from the stationary distribution. That might help people to figure out what to focus on if they are playing with operators to increase the global ESS values.