Band Recovery ?'s

[branta]

Hi,
I am new at band analysis and to Program MARK, so I apologize in advance if these questions seem elementary. I have read much of the “Gentle Introduction Manual” and it has helped greatly, but I still have a few questions.
I am trying to analyze band recovery data for Canada geese banded 4 years (k=4) and with 5 years of recovery (/=5) and I have 2 age classes (adult and young). I imported the following recovery matrix and selected the Brownie et al. recoveries.

recovery matrix group = 1;
42 23 14 10 7;
39 23 20 11;
75 45 23;
80 45;
0;
316 235 504 461 0;
recovery matrix group = 2;
107 85 30 16 11;
100 58 38 29;
77 71 32;
129 54;
0;
694 521 545 563 0;

I ran 4 models: CJS model and the 3 models used in Chapter 10 which were S(a*t)f(a*t), S(a) f(a*t), and S(a) f(a). If I need more models or used the wrong models, please comment.

My next question concerns some large standard errors from the estimates. When I ran the S(a*t)f(a*t) model I got real large standard errors (80-162) for just 4 estimates (4:s, 8:s, 13:f, and 18:f). The same thing happened for estimate 12:f when I ran the S(a) f(a*t) model. I removed the 5th year of recoveries to see if the problem would still occur, and then the SE were normal. I retyped my data, reran, and still have the same problem. How do I fix this or what is occurring?

My other questions regard which model to select. Which model do I perform the bootstrap GOF test on? Then after I have made the c-hat adjustment, do I use the LRT test to pick among the models that have the highest AIC values?
Finally, is there any other analysis that I need to do besides what I have mentioned above for these estimated survival and recovery rates?

Any comments would be greatly appreciated.
Thanks,
Bob

[cooch]

I am trying to analyze band recovery data for Canada geese banded 4 years (k=4) and with 5 years of recovery (/=5) and I have 2 age classes (adult and young). I imported the following recovery matrix and selected the Brownie et al. recoveries.

recovery matrix group = 1;
42 23 14 10 7;
39 23 20 11;
75 45 23;
80 45;
0;
316 235 504 461 0;
recovery matrix group = 2;
107 85 30 16 11;
100 58 38 29;
77 71 32;
129 54;
0;
694 521 545 563 0;

Which of the two recovery matrices corresponds to 'marked as adults', and 'marked as young'?

I ran 4 models: CJS model and the 3 models used in Chapter 10 which were S(a*t)f(a*t), S(a) f(a*t), and S(a) f(a). If I need more models or used the wrong models, please comment.

To some degree, you're missing the point - the models you run (i.e., the candidate model set) should comprise models that are of interest to you - for either biological or management reasons. The models listed in Chapter 10 are only intended to provide some guidance on how to build the models, not to make specific suggestions on what models to fit. Selecting models is something you need to think hard about on your own...

My other questions regard which model to select. Which model do I perform the bootstrap GOF test on?

As noted in detail in Chapter 5, you run the GOF on the most general model in your candidate model set...

Then after I have made the c-hat adjustment, do I use the LRT test to pick among the models that have the highest AIC values?

No - have a look at Chapter 4. LRT and AIC represent different model selection paradigms. Pick one or the other, but don't mix and match.

[branta]

The recovery matrix group=1 is for adults and group 2 is for young.
Thanks.

[cooch]

When l <> k, not everything is identifiable (leading to the problems with SE's you noticed). This is explained in fair detail in the Brownie et al. handbook, and more accessibly in the Williams, Nichols & Conroy book. I'd suggest getting a copy of the latter, and looking there.