New RMark versiopn

[jlaake]

I posted a new version of RMark. It requires R2.12.0 or higher. Note that with this version of R, a single package works for both 32 and 64 bit so there is no longer a separate 64 bit version. For linux, download the new source (not Windows binary) zip file and INSTALL the package. To do that you'll need to update R and Rtools. If installation from the zip file does not work, unzip into a directory and then use the r cmd INSTALL RMark command from the directory containng the RMark package. There are just a few minor fixes and an example of the CRDMS model that Andrew Paul created.

On the RMark phidot page we have acrhived all of the external (outside R) documentation for RMark in a single zip file. I added a pdf that shows all models in MARK (at the time Gary sent me list), the ones supported by RMark and the example data file in RMark if there is one, and the names of all the parameters in the model. I'll keep this updated as I add models supported by RMark. It has rather tiny print to fit each line on a page, so you'll want to use the zoom in Adobe reader.

[pmm]

Is there a 64 bit version of mark.exe as well, or will Mark have the same address limits the 32 bit RMark did?

[gwhite]

There is a 64-bit version of the mark.exe file, but I doubt that you really need it. Evan has it up on euler, and I can provide for other machines if necessary. However, I've never heard of anybody running out of memory with mark.exe (unless they did something very arcane), but rather CPU time is the limiting factor.

Gary

[jlaake]

Gary-

I chatted with him offlist and the problem isn't memory it is CPU time. He is trying to fit a model with 400K rows of data and individual covariates. The model ran for 4 days before he terminated the job. I suggested that he either accumulate capture histories with the same covariate value or try some code I developed that is faster for individual covariate models. It works well with my 10K data set but 400K maybe a challenge.

--jeff

[pmm]

Thanks!

I am still experimenting with this and will post some findings if I reach a good solution.

-Patrick McCann