Sample Size Requirement for MARK

[emlehmer]

I am examining trans-seasonal survival from a mark-recapture study and was asked by a reviewer to use MARK to analyze these data. However, my sample contains only about 65 animals (65 recaptures over 4 seasons out of 1300 total captures during this period). Is this a large enough data set to use MARK? Thank you!

[cooch]

I am examining trans-seasonal survival from a mark-recapture study and was asked by a reviewer to use MARK to analyze these data. However, my sample contains only about 65 animals (65 recaptures over 4 seasons out of 1300 total captures during this period). Is this a large enough data set to use MARK? Thank you!

Well, if you see the 65 individuals with high probability, then 65 is a lot!

Basically, the 'big law' in mark-recapture is 'do whatever you can to increase probability of detection'. There are studies with <65 animals that do just fine, because p is very high (say, >0.7). On the other hand, I frequently review papers where p<0.1, depsite having sample sizes that are very large (in some cases, ten's of thousands), and generally 'disappoint' the authors when I point out that with a detection probability that low, there will often be real limits to the strength (and complexity) of inference they can draw. Its amusing to see the most parsimonious model for a sample as large as this be a dot model.

But, it sounds as if in your case the detection rate is pretty low - and it may not actually be possible to do much of anything with your data - at least, nothing particularly fancy. I'd suggest doing a simple test with some simple models, to get a rough idea what p is for your data. 65 out of 1300 is a very small fraction, but perhaps (as noted above) the 65 are from a particular set of individuals.