Fixing parameters during years with no recaptures/recoveries

questions concerning analysis/theory using program MARK

Fixing parameters during years with no recaptures/recoveries

Postby jbauder » Fri Dec 06, 2019 8:50 am

Hello,

I am using a multi-state model to estimate cause-specific mortality in black bears using a 38 year data set. Our states are alive, dead from harvest, and dead from all other mortality causes, with transition probabilities between dead states fixed to zero. I am estimating parameters for live recaptures (pA), dead recoveries from “other” mortality sources (rO), probability of mortality from harvest (PsiH), and probability of mortality from “other” sources (PsiO). I fixed probability of recovery from harvest mortality (rH) to one because all hunters are legally required to report their harvests.

The data are fairly sparse and many years have no live recapture, no dead recoveries from “other” mortality sources, and even a few years with no reported harvests of our marked bears (even though there was a legal harvest season and unmarked bears were harvested in every year). I am not considering a fully time-dependent model (e.g., pA(time)) for that reason (i.e., most years have real parameter estimates virtually zero and “wonky” SEs) but will account for time-effects by creating time bins in RMark (e.g., 1979-1990, 1991-2005, 2005-2016) or using polynomial effects of time (e.g., Time+Time^2). I am wondering, however, if I should still fix parameters equal to zero for years in which no recaptures, recoveries, or harvests occurred. For example, if years 3, 9, 10, 11, 12, 13, 20, and 27 had no live recaptures, should I fix pA=0 for those years even if the model is not time-dependent (e.g., a sex*age model or a sex*Time^2 model)? It seems tricky with regards to harvest mortality (PsiH) because we know there was a non-zero probability of harvest in each year (i.e., an open season) but none of our marked bears happened to be harvested in some years. Does it makes sense to set PsiH=0 for those years when the model structure is not fully time-dependent? Or should parameters only be fixed for a given year when using a fully time-dependent structure?

More generally, does having years with no recaptures constitute a biologically justifiable reason to fix recapture to zero? Or is it better to forgo the fully time-dependent structure for simpler models? That’s a scenario I seem to encounter in all my mark-recapture data sets!

Thanks,
Javan
jbauder
 
Posts: 56
Joined: Wed May 25, 2011 12:01 pm

Re: Fixing parameters during years with no recaptures/recove

Postby jlaake » Fri Dec 06, 2019 6:16 pm

Unless there was no sampling effort or no harvesting, I don't see a good reason to fix parameters to zero just because there were no observations, especially if you have small sample sizes. I suggest you stick with simpler models. My 2 cents.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Fixing parameters during years with no recaptures/recove

Postby jbauder » Fri Dec 06, 2019 8:30 pm

Thanks Jeff, that makes sense!
jbauder
 
Posts: 56
Joined: Wed May 25, 2011 12:01 pm


Return to analysis help

Who is online

Users browsing this forum: Google [Bot] and 9 guests