Statistical Modeling, Causal Inference, and Social Science: bugs.R question

bugs.R question

This one's just for the bugs.R users out there . . .

Boris S.

I'd like to second Andrew's comment about looping within a WinBugs model. I was previously invoking Bugs.R repeatedly in an R loop (for a Monte Carlo analysis), and it was painfully slow, as the fixed cost of invoking Bugs.R and WinBugs is relatively high. But after putting the loop inside the model, things sped up considerably. But do note that this requires lots of memory ...

Hui

Putting the loop inside the model works, and as Boris pointed out, it did speed up. However one consequence is that one would not be able to obtain some statistics, like DIC, for each indiviual loop. We might only be able to obtain DIC for the whole model, i.e. for all the individual analysis together.

Andrew

Hui,

Yes, you'll just have to compute DIC by postprocessing the simulations, writing a little R script.