CV and Var of N in POPAN

[reisinger]

Forgive me if I've overlooked this in the MARK book, or if its a basic statistical calculation...

In POPAN, CV and Var are not reported for N. How can these descriptors be calculated/obtained?

[cschwarz@stat.sfu.ca]

Estimates of se and N are available in the derived parameter area.

The var(N-hat) = se**2.
The cv(N-hat) [more modern usage is relative se rather than cv] is computed as se(N-hat)/N-hat.

[ehileman]

Hi all,

I have what I assume to be a trivial but related question, so I thought I'd revive this thread. I am using model averaging in Popan to estimate the derived N_i (i.e., N-hat). I used a CJS GOF test to provide a crude estimate of c-hat for my Popan global model. However, I read in the Popan derived output that N-hat (along with all other derived estimates) are not corrected for c-hat. I assume this means that the N-hat SE and CIs used a c-hat of 1?

Anyway, I want to calculate adjusted c-hat SEs and CIs for all five of my derived N_i estimates. Can I adjust the SE doing this: SE = sqrt((adjusted c-hat)*(variance))? For the CIs, is the equation described in

14.10.1. Estimating CI for model averaged abundance estimates

of "the book" appropriate here given that this is an open population estimator?

Many thanks for your help!

Eric

[cooch]

Hi all,

I have what I assume to be a trivial but related question, so I thought I'd revive this thread. I am using model averaging in Popan to estimate the derived N_i (i.e., N-hat). I used a CJS GOF test to provide a crude estimate of c-hat for my Popan global model. However, I read in the Popan derived output that N-hat (along with all other derived estimates) are not corrected for c-hat. I assume this means that the N-hat SE and CIs used a c-hat of 1?

Anyway, I want to calculate adjusted c-hat SEs and CIs for all five of my derived N_i estimates. Can I adjust the SE doing this: SE = sqrt((adjusted c-hat)*(variance))? For the CIs, is the equation described in

14.10.1. Estimating CI for model averaged abundance estimates

of "the book" appropriate here given that this is an open population estimator?

Many thanks for your help!

Eric

Yes , use the same approach, for the same reasons.

[ehileman]

Yes , use the same approach, for the same reasons.

Thanks for the reply, Evan!

One quick follow up. Regarding calculating confidence intervals, am I correct in thinking that to make this method appropriate for Popan N-hat estimates I should set (f0) = Ni - Mt (rather than Mt +1), where Ni is the population estimate at time i and Mt is the unique number of individuals captures at time t? Using Mt+1 in this situation doesn't make any sense to me and would result in unreasonably wide CIs.

Eric

[cooch]

Yes , use the same approach, for the same reasons.

Thanks for the reply, Evan!

One quick follow up. Regarding calculating confidence intervals, am I correct in thinking that to make this method appropriate for Popan N-hat estimates I should set (f0) = Ni - Mt (rather than Mt +1), where Ni is the population estimate at time i and Mt is the unique number of individuals captures at time t? Using Mt+1 in this situation doesn't make any sense to me and would result in unreasonably wide CIs.

Eric

Seems reasonable. Don't think there is a definitive answer, since there has been little application of model averaging to open population abundance estimates, since (in general) they're pretty lousy (i.e., imprecise).

[ehileman]

Seems reasonable. Don't think there is a definitive answer, since there has been little application of model averaging to open population abundance estimates, since (in general) they're pretty lousy (i.e., imprecise).

Evan,

Thank you for your insight and prompt reply. If the dataset I'm currently working with adequately met the assumptions of closure, I would have opted to use closed models instead. Unfortunately, they don't.

Cheers!

Eric