search

xfim / ggmcmc

Graphical tools for analyzing Markov Chain Monte Carlo simulations from Bayesian inference
111 stars 31 forks source link

Fix: issue #18 simple STAN test case #20

Closed jucor closed 11 years ago

jucor commented 11 years ago

In issue #18, @dmenne found that importing from STAN fails when only the chain has only one parameter. This pull-request fixes it.

If errors subsist, I recommend merging pull-request #17 that uses reshape2, which will fix any problem with the melt() used in ggs_chain().

dmenne commented 11 years ago

The error is gone, but ggs_autocorrelation(S) creates warning messages: I suggest to make "non-parallel backend" the default, everything else is likely to lead into problems. As Kernighan stated: first make it right, then make it fast. This package is still terribly buggy. Thanks a lot for your efforts. I think you should ask xfim for full access, as far I heard he is very busy with other things.

Warning messages: 1: In setup_parallel() : No parallel backend registered 2: In 2:nLags : numerical expression has 1000 elements: only the first used 3: In 1:nLags : numerical expression has 1000 elements: only the first used

dmenne commented 11 years ago

Note that this also fixed issue #14 which I thought was a bogus one, but I did not re-check the case of 1 variable. The decision to drop by default was one of the worst decisions in early R development.