[R] MCMC tools
Therneau, Terry M., Ph.D.
therne@u @end|ng |rom m@yo@edu
Tue Dec 27 21:32:14 CET 2022
Is there a convenient package that computes standard covergence summaries for and MCMC
run? This is something that I likely knew once and have now forgotton.
More detail: I'm trying to understand the MCMC done by a particular model called
Subtype and Stage Inference (SuStaIn), suffice it to say that I am cautious of some
details. The model doesn't fit into the standard packages, so I've set it up and run my
own Metropolis chains. I don't want to expend energy also creating R-hat, ESS, and other
sensible summaries; even more so to find the inevitable programming mistakes I'll make if
I create them myself.
For the truly curious. I have measurments of tau depostion for 86 brain regions (43 *
left/right) of interest from 1925 scans. One hypothesis in dementia research is that tau
deposition occurs over time, in a pattern; there is likely more than one pattern. The
algorithm is looking a permutations of the 86 regions, in search of a small collection
that best fits all the subjects. There is a general background of statistical work that
has shown that ranking is a hard problem, in the sense of having a small variance for
individual ranks. SuStaIn tends to give small variances.
Terry M Therneau, PhD
Department of Quantitative Health Sciences
therneau using mayo.edu
[[alternative HTML version deleted]]
More information about the R-help