unexpectedly high survival estimates

[donna]

Hi all,

I'm conducting mark-recapture analysis in RMark for a dataset on woodland birds. I have divided the dataset into individual species, and am running Barker models on colour-banded species, and CJS models on the rest.

For some species, the survival estimates (using both Barker and CJS models) seem far too high (1 or close to 1) based on my recapture rate. Some datasets are small (10-30 ch), but others are over 100 ch. I've gone through my dataset in full, and nothing stands out as strange or different about the species with unexpectedly high survival estimates.

Note that my nocc = 3. I thought this might be a contributing factor, but it doesn't really explain why some species have reasonable-looking survival estimates with low recapture rates while others are seemingly off the charts.

Would appreciate any insights.

Thanks so much!

[jlaake]

With 3 occasions, unless p is very high I would expect this and the reason there is no pattern is likely due to randomness. My guess is that your precision is poor and confidence intervals are very wide. I would suggest rather than splitting your dataset by species that you combine them and try to gain support for parameters across species. You can use species as a group variable and then examine species differences with AIC although with the small sample sizes (100 is small with only 3 occasions) you will likely not find any differences.

--jeff

[donna]

Thanks very much, Jeff, that's really helpful to know. I'll try as you suggest