General analysis and use of survival estimates

[murray.efford]

As long as you know the locations of all stations (including those that did not catch an animal) it is possible to go ahead with a single-occasion point-based spatial analysis (detector type 'count' in secr). Ideally you would also allow for varying effort between sites (a trap-level covariate), but if the variation is 'small' it may have a small effect on the estimates: this could be checked by modelling with and without such an effect at the site where you have the data, and by simulation (not as hard as it sounds).

[As an aside - you might abandon point-based spatial analysis and treat the data as if they resulted from area searches (Royle & Young 2008 and Efford Ecology in press), but this is a rough approximation unless many stations were used in each year (so that capture effort was more or less even across the study area up to a definite boundary). Would be interesting to check with some simulations how well this would work.]

I'm sorry this is all a bit vague - a lot depends on the data, so I would try it and see.
Murray

[mldavis13]

Hi Jeff (if you're still reading this) or anyone else who may be following,

Concerning the problems of individual covariates in JS model formulations. According to my understanding of Ch. 13 "covariates" which don't change over the study period - such as sex groups are less problematic, but those that do such as age can be a bigger issue. Recent advances in this area seem to go the way of Bayesian statistics (Dupuis and Schwarz 2007; Schofield et al. 2011). Short of going Bayesian or leaving program MARK for a SECR or DENSITY approach, does including an age-structure (differences in survival among age groups in my species are large and significant) in my analysis mean that JS is not possible and I am limited to the CJS approach?

Thanks for the help. I am currently also looking into SECR as an option but would like to clarify this point!

Miranda

[jlaake]

Short of going Bayesian or leaving program MARK for a SECR or DENSITY approach, does including an age-structure (differences in survival among age groups in my species are large and significant) in my analysis mean that JS is not possible and I am limited to the CJS approach?

Yes, that is what I tried to convey in my emails. The problem is time-varying animal-specific covariates. Age is one of those with a know transition given you know the initial age of everyone. You can have age-effects only if the only entrants into the population are all the same age -- for example, young of the year. That would be a cohort-specific analysis and you would be estimating the initial size of each cohort and then its survival followed through time. The popn in a year is the sum across cohorts.

--jeff