Applying MARK estimates to DENSITY

[sixtystrat]

Is there a way to apply my MARK estimates of N to estimate density using program DENSITY? I used a robust-design model structure in MARK with different sex-time interactions and individual covariates, etc. and am having trouble duplicating that structure in DENSITY. I and would like to take full advantage of those MARK estimates when estimating density. Thanks for any help.
Joe

[murray.efford]

Joe
I think the direct answer to your question is 'no' for two reasons: (i) N does not appear in spatially explicit capture-recapture models (certainly not the N estimated by MARK) (ii) in Density I knowingly made it hard to fit models with time-specific detection because these usually involve superfluous nuisance parameters. However, you should be able to fit a model like that with the recent beta release of the R package 'secr' e.g. secr.fit(yourdata, CL = T, model=list(g0 ~ session + t * g + zi), groups = 'sex') where 'sex' and 'zi' are respectively categorical (factor) and continuous individual covariates. As it stands, this may be an over-night job, and I suspect AIC or a score test would show the t*g interaction is way over the top. Check out www.otago.ac.nz/density/SECRinR.html.
Murray

PS Depending on your application, it may be more interesting to use the full likelihood (CL = F) and model differences in density between sessions while doing without the continuous individual covariate (only available with conditional likelihood).

[murray.efford]

Correcting my rushed reply from yesterday: 'secr' has no need of groups when fitting conditional-likelihood models, so the model should be: secr.fit(yourdata, CL = T, model=list(g0 ~ session + t * sex + zi)). (Categorical and continuous individual covariates can both appear by name in the model. If you care about sex-specific values of the real parameters they can be retrieved from the fitted model with 'predict' and/or 'derived')
Murray