www.phidot.org

Dear list,

I have been unsuccessfully trying to assess model convergence after running MCMC on a multistate model in MARK. I used the following code in R:

mcmcdata=create.mark.mcmc(filename,ncovs,nmeans,ndesigns,nsigmas,nrhos,nlogit,include=F);

and got

Error in xmcmc[i, ] = readBin(z, numeric(), n = nvals) :
number of items to replace is not a multiple of replacement length

By debugging the create.mark.mcmc I noticed two things:
The (while) loop in create.mark.mcmc causes the error at i=9988, just before reaching the 10000th sample.
Secondly, within the xmcmc object, values range between -1.797342e+308 and 1.797275e+308, yet the editor output of the mcmc simulations seems to be providing reasonable estimates of parameters.

I used MARK 6.1 on Window NT (x64), and tried both the 32 and 64 bit versions of R.

Anyone any ideas?

Thanks in advance,
Best wishes

Caspar

This is not part of RMark and should be listed on the Analysis Help sub-forum. When you get that type of result it can be a disparity between binary lengths - 32 vs 64 bits.