RMark vs. MARK model likelihood error

[jwarren]

When I append model results from RMark to MARK for running GOF tests I receive the following error message: "Have you changed the data, because the saturated model's likelihood has changed!". I also noticed that the model deviance in MARK is considerably lower than the deviance in RMark for the same model. Any ideas/assistance would be greatly appreciated.

[dhewitt]

I just dealt with this same issue, I think. Are you using both export.chdata() and export.model() to get both the input file and the model results for appending in MARK? Chances are you need to because the original INP you imported into RMark may not be appropriate for building the MARK project to go with the RMark model. If you're using the current R and RMark versions, export.chdata() won't work right now (which led me to just use my original INP to build the MARK project, but that doesn't work). Jeff Laake just patched the function for me, so if this is your problem you might contact him for it (or I can pass it along).

The reason is that the ordering of groups is set by different defaults in RMark than MARK, so you need to export the processed data from RMark to pull into MARK so that the INP matches the model structure.

If this isn't your problem, I wasted a lot of keystrokes.

[jwarren]

You're exactly right in how I'm using the export.chdata() and export.model() functions. Thanks for the help, I'll check in with Jeff Laake.