Accessing MCMC chain values/real estimates

[FishyAC85]

Moved to Software problems subforum. Under same subject

[jlaake]

MARK is composed of a Windows interface and a FORTRAN program MARK.EXE. RMark can only use the FORTRAN code. It is not well known but some aspects of MARK like MCMC are implemented in the Windows interface and are not available to RMark. This post should be in the MARK software problems sub-forum.

[jlaake]

I emailed Gary and it turns out I was wrong about MCMC. It is implemented in the FORTRAN code but apparently there are several screens of information that need to be completed. No promises but I may look into it. I can see how you might think that RMark would handle it because of the function create.mark.mcmc which I included to help pass MCMC results to coda package.

[FishyAC85]

Thanks Jeff. Any ideas to access them would be helpful. Dealing with a reviewer that wants to see chain values.

[FishyAC85]

In case anyone else ever wants the chain values, I wanted to update RE the solution.

As a reminder, I was running a multistate model using MCMC estimation, which is not supported directly in RMark.

1) You must run the model in RMark, use export.MARK, and then import to Program MARK. My model was very simple so I had just been making and running it in Program MARK and this was creating an issue when I tried to bring the .BIN file into RMark/coda to read.

2) The output file created when you run your model in Program MARK (.tmp) contains a section that lists the information you need to run the create.mark.mcmc command to open the .BIN file containing the chain values into R. Just search for something like ncovs in the tmp file and you'll find what you need.

The .BIN file opens as a mcmc.list file type but I could change it to a matrix (as.matrix) and then do what I needed to with it.