Issues when rewrite winbugs code to stan

General
3

Thanks for looking into this. There are many strange things:

  1. I am struggle with why should I assign the mu2, mu3, mu4 …, tau3, tau4, tau5… a different name for mu and tau. Can I do mu[j] and tau[j]? maybe not, because after I do that, there are even more issues. Not sure why?

  2. The results from stan is different from the winbugs result, do you know why? Is that what I did is not correct?

                                      Stan results:           Winbugs results:

tau 3 ( 95% CI) 0.39(0.02,1.59) 0.32 (0.01,1.5)
tau 4 ( 95% CI) 0.57(0.6,2.21) 0.54(0.01,2.25)
tau 5 ( 95% CI) 0.71(0.8,2.91) 0.75（0.09，2.82）

  1. if I did what is your suggestions from other post like below: vector<offset=mu3,multiplier=tau3>[Ntotal] theta3;
    vector<offset=mu4,multiplier=tau4>[Ntotal] theta4;
    vector<offset=mu5,multiplier=tau5>[Ntotal] theta5;

There are many more issues if I only use adapt_delta = 0.99, so that is what I concern about most.

But as you know after I add the control = list(adapt_delta = 0.999,max_treedepth = 15) there are no issues. Is this to say, we should always use those options? I do not personally think we should always use this kind of options, do we?