www.phidot.org

I am trying to run a Pradel and Seniority (Pradsen) model in RMARK.

When I run the constant model for all parameters, I get the following error in R:
Error in if (x4 > x2) { : argument is of length zero
********Following model failed to run : Phi(~1)p(~1)Gamma(~1) **********

And the following error in the MARK output:
ERROR -- Numerical convergence never reached.

I get estimates with no errors if I:
1. Run either a CJS or Pradel model in RMARK with a constant model for all parameters
2. Force a sine link for the Pradsen model in RMARK with a constant model for all parameters
3. Run a constant model for all parameters in MARK

Can someone help me troubleshoot what is going on? I can send along the code if necessary.
Thanks in advance for your help!
Gina

Gina-

Send me your code and data file and I'll look through it. jeff.laake@noaa.gov

--jeff

Sorry, this is my problem. I have the wrong default link function for gamma. It is currently using log link instead of logit link. I'm going to have to do some detective work in my archive because I fixed this problem in Sept 2009 with version 1.9.3. Make the following small change and it will work:

Gamma.dot=list(formula=~1,link="logit")

I'll fix this in the next version but in the meantime the above will work to stop it from using the log link. Whenever you see the error message you got, it is best to go look at the MARK output file in the working directory and see if you can work out what the problem is. RMark is creating an input file for MARK so you can compare output files from MARK interface and RMark interface and look for differences. If you look at the output file that was created from RMark model that it specifies Parm-specific links and then shows logit for Phi and p and log for gamma. Use the above to force a logit link for gamma and it will work.

--jeff