www.phidot.org

Dear all,

I have been running some CJS models using RMark on a dataset with 8 groups (sex+4 age groups) and 22 occasions. Program release (called through R) has been giving good (non-significant) test results and my median-c estimation has been running fine as well.

I recently tried "truncating the dataset" to exclude some years for which some temporal covariates of interest are unavailable. The truncated dataset has 18 occasions. With this dataset release gives the following error:

> release.gof(grimch.pr)
Set the parameter MAXOCC larger
RELEAS ERROR TERMINATION
Chi.square df P
TEST2 107.0017 49 0.0000
TEST3 45.1045 65 0.9715
Total 152.1062 114 0.0099

In the extended output (at the very end) it similarly says:
"
TEST 1.R16: Test of Phi(16)p(17) equal across groups,
assuming higher order parameters are equal across groups.

* * WARNING * * Chi-square table calculator dimension error.
INPUT --- proc stop;

Set the parameter MAXOCC larger"

I have searched online, in the MARK book, and the forum for "MAXOCC" and various versions of it to little effect. On the page: http://csprousers.org/downloads/cspro4. ... enotes.txt

I found this statement:
"MAXOCC function. This new function returns the maximum number of
multiply occurring records or the maximum number of multiply-occurring
items defined for a group in the dictionary."

but I'm afraid I don't understand what this means.

I'm guessing that it is due to my dataset having too few occasions to fit the model appropriately. Could some please confirm that this is the meaning of this message? Or is there a MAXOCC parameter I can alter by running Release manually that could solve this problem?

Thanks,

Miranda

I could be wrong, but I doubt that Gary is using a function from the Census Bureau software in MARK. I think MAXOCC is just the name for a variable used in the RELEASE code.

My suspicion here is that in truncating your data set you have left in something in your INP or setup code in R/RMark that makes it think it has more occasions that you actually do. Without more info, we can't be much more help unless someone out there has encountered the same problem.

- Dave Hewitt

I believe MAXOCC stands for maximum occasions. I don't believe there is any way to set this for RELEASE but Gary White would be the person to address that. If you want to run manually, you can use export.MARK in R and then import into MARK interface and run RELEASE from the MARK menu. I'd be curious if you got anything different.

regards --jeff

jlaake wrote:I believe MAXOCC stands for maximum occasions. I don't believe there is any way to set this for RELEASE but Gary White would be the person to address that. If you want to run manually, you can use export.MARK in R and then import into MARK interface and run RELEASE from the MARK menu. I'd be curious if you got anything different.

regards --jeff

That would be correct. You could change that only if compiling from raw source code.

I'd concur with Jeff's suggestion -- pull it in through the MARK interface, or, simply run it standalone from the command line (full details in Chapter 5).

Thanks for the suggestions! I am currently looking into the options you suggested (and error-checking my code) so may be a while before I get anywhere, but just wanted to say thanks for the input!

Miranda

I ended up running into this error. I queried Gary and he found a place where the code had a fairly low restriction and increased the array space. I no longer get the error with my examples. I suggest that you go to Gary's mark website and download the update.zip and copy the files into the location where you have the mark program. Rel32.exe is the release code for gof. I found that even though I was getting the error it was still computing the test2+3 values and the error came at the end of the output file.--jeff

jlaake wrote:I ended up running into this error. I queried Gary and he found a place where the code had a fairly low restriction and increased the array space. I no longer get the error with my examples. I suggest that you go to Gary's mark website and download the update.zip and copy the files into the location where you have the mark program. Rel32.exe is the release code for gof. I found that even though I was getting the error it was still computing the test2+3 values and the error came at the end of the output file.--jeff

Easier to simply re-install MARK with the latest build -- rel32.exe has been updated in that build, I believe.

Thank you both for the assistance!
M