Problems with running mark-resight models

[MWhitehead]

Hello!

I have been experiencing some difficulties running an immi-emi logi-normal for mark-resight analysis.

After quadruple checking my .inp file I discovered the IELNE_IndividualMarks.inp example file and I run into exactly the same problem using this.

After setting up the analysis (12 occasions/3 primary/4 secondary) and selecting a general all parameters constant model (=15 parameters), with either SIN or LOGIT link function, I get the following error message when I click OK to run and the DOS window leaps into action:

forrtl: severe (161): Program Exception - array bounds exceeded
Image PC Routine Line Source
MARK.EXE 0065AFE6 SETLINK 184 setlink.for

I am running the most current version of MARK (2/6/11) on a PC in Windows 7. I have changed compatibility of MARK_INT.exe to WinXP with no joy for my analysis. I can run simple closed pop analyses fine.

Can anyone shed some light as to what's going wrong here?

Thank you in advance.

Mike

[bmcclintock]

Hi Mike,

This hasn't come up before, but I believe this error is related to setting the alpha parameter constant. The alpha parameters for each primary occasion are constrained to sum to 0, so I'm guessing that setting them constant in the PIMs is causing a problem. If you really want to run a model with constant alpha's, then don't set them as constant in the PIMs. Instead, leave them as time-dependent in the PIMs and fix them all to 0 from the run window.

Cheers,
Brett

[gwhite]

I just put a new copy of MARK on my web page that fixes the problem of setting the alphas constant in the PIMs. However, such a model is not biologically realistic, in that you are assuming all the estimated alphas are the same, and the last one is equal to the negative of the sum of the estimated (constant) values.

Gary

[cooch]

New version mirrored here also:

http://www.phidot.org/software/mark/download/

[bmcclintock]

Following up on Gary's comment, one might set the alpha's constant and fix them to 0 to assess the evidence for or against geographic closure (although one should already have a pretty good handle on this based on the marked population's movements).

Alternatively, one could use this to "trick" the immigration/emigration logit normal estimator to behave much like the regular logit normal model, but one which allows marks to be introduced within primary sampling occasions. For this trick, the number of marked animals in the superpopulation would be set equal to the number of marked animals that are actually within the study area for each secondary occasion (i.e., T_{ij}=M_{ij} using notation of The Book), Nstar would be set equal to Nbar (via PIM or design matrix), and the alpha's would be set constant and fixed to 0.

Brett

[MWhitehead]

Thank you all for the prompt replies.

I guess my ignorance was showing in that I was trying to get my data up and going with an unrealistic model.

I'll keep trying.