Iguana Analysis

#1
Hello!

I am a student currently working on iguana data from an island in the Caribbean close to Cuba (dating 20 years). It is a capture-recapture study as well, although not strictly, where every time we go to collect data we find old recruits (recapture) and some new ones (newly captured). The information gathered is

  • Dates of Capture/Recapture
  • Iguana IDs
  • Weight
  • Length (from head to just above base of the tail)

I have done a logistic analysis previously on the female iguanas before, studying their birth rate based on location of the island and the total amount of eggs.

I would ultimately like to do a survival analysis, but don't really have an idea on how to go about this. I heard of Generalized Additive Models (GAMs), but don't know how to incorporate the capture-recapture aspect into it. I would greatly appreciate some help as to whether what methods I should try to use, or if GAMs is the way to go, how to go about using them for this study. I am not opposed to purely classical, bayesian, or mixed of the two. Thanks in advance!

~ris
 

bugman

Super Moderator
#2
Though mark repature analysis is not something I am familar with, I will say that if you have a mixture of binary and continuous varaibles, then yes GAMs will be useful to you.

I dont know how to incorporate the mark-recapture component into this analysis (I haven't used this method before), however I have attached two PDFS written by an ex-collegue.

She specialises in mark-recapture analyses and these papers are likley to be very useful.

P
 
#3
Thanks a lot for the papers! It's really nice to have such resources from experts on the topic!

So while looking through on different papers on the subject, a lot has mentioned the R package R capture. Considering that I have the dates of capture, and not really sessions, I decided to group the dates by year (not one iguana was captured more than once in one year), so my periods are now years. I designed a matrix of the iguanas versus period of capture for a total of 16 periods. So far I am not incorporating length nor weight, but could these possibly be used as some sort of prior for survival rates in a Bayesian framework?

So far there aren't any serious issues (that I know of), except some models leading to singularity issues for model fitting, but I just want to make sure I am not going in the wrong direction with this. Any possible precautions or assumptions I should take into consideration as I use Rcapture? Is there some risk involved by directly converting actual dates of capture into groups of capture periods by years? Thanks in advance!

~ris