confirmation on how to interpret MCMCglmm summary output

Question

i want to know if post.mean is similar with estimate result that we see in glm model or different? thus pMCMC can be consider as p-value? i have not found people really explain on summary analysis usually they focus on something else.

lastly, how to set prior and how i can be sure that my prior is correct? here link i have read but still cannot understand. http://www.wildanimalmodels.org/tiki-download_wiki_attachment.php?attId=24 https://cran.r-project.org/web/packages/MCMCglmm/vignettes/CourseNotes.pdf this my result that i obtain.

 Thinning interval  = 10
 Sample size  = 1000 

 DIC: 0.3723323 

 G-structure:  ~time

     post.mean  l-95% CI u-95% CI eff.samp
time     11189 4.032e-17    13211     1000

 R-structure:  ~units

      post.mean l-95% CI u-95% CI eff.samp
units    308299    24934   815369     68.1

 Location effects: sleeve ~ pain + stiff + diff 

            post.mean l-95% CI u-95% CI eff.samp pMCMC
(Intercept)    72.812 -334.677  433.864    467.2 0.550
pain          -33.016 -156.926   68.739    451.5 0.522
stiff           1.016  -95.617   88.061    479.0 0.984
diff            1.799  -25.380   25.683    772.7 0.902```

score 0 · Answer 1 · answered Dec 03 '19 at 21:36

Yes, the pMCMC is a p-value, however it does not appear that your MCMC is converging. You should define a number of burn-in samples to allow the chains to converge. For instance, your variable stiff is estimated to be -95 < stiff < 88, and it is only being effectively sampled < 500 times.

As for your other question, there's no way to confirm whether your priors are defined correctly without seeing how they are defined, and knowing more about the problem you're trying to solve.

confirmation on how to interpret MCMCglmm summary output

1 Answers1