help running Jolly-Seber

[mike]

Hello, new to using MARK, I have checked the FAQ and can not find help so I'm asking with a new post. I'm trying to run a JS and keep getting a "numerical convergence never reached message, exit code 3/4 a30.

The data are set up in sampling events horizontally and unique tag numbers are the row and either 0 or 1 for each sampling event. I have even tried a fake data set with only 6 events with the same results.

In the specification window I am selecting Jolly-Seber and setting the encounter occasions to 6. Once the survival window opens I select run current model and name the model "Jolly-Seber” then hit run and that’s when the program stops running and I get the non convergence message.
Thanks in advance for any help.
Mike

[gwhite]

Mike:
The Jolly-Seber model data type that you are using has a lot of difficulty converging, even though I'm 99.9% sure the model is correctly implemented. Alternatives if you don't need to know the population size are the Pradel models, or the Link-Barker model (accessible as a change in data type under a Pradel model). If you need to know the population size, then try the POPAN model. It converges much better than the JS data type. The reason that POPAN is parameterized with the probability of entries is because this makes optimization and convergence easier. Just be sure to use the multinomial logit link, available under the user-specified link option.
Gary

[cschwarz@stat.sfu.ca]

Speaking of using POPAN to run Jolly-Seber models....

I'm in the middle of writing a chapter for the Gentle Introduction to MARK on using the POPAN module. It has screen shots of the steps necessary and what to watch out for when using POPAN.'

Not yet completed, but people are welcome to look at the current stuff. Follow the URL
http://www.stat.sfu.ca/~cschwarz/POPAN/ ... ndance.pdf

The first two pages of the pdf file are blank, but it does continue later on.

[mike]

Thank both of you for the help. I have another question but will wait until I read Carl's instructions before I ask. If I may I'd like to explain my data and see if you think the analysis is proper. I have 3 yrs of hook-n-line mark-recapture data on Arctic Grayling, I had a total of 46 sampling weeks (18,16,12, weeks). I orginally began by treating each week as an event in the capture history file. I was unable to run POPAN on this data so I then treated each year as an event and coded the history file by groups (i.e. 101 142;) this worked. I'm wondering if it was correct?
Again, thanks for the help.