chat.nk application
41 views
Skip to first unread message
jcl...@utk.edu
May 16, 2023, 11:41:02 AM5/16/23
to secr
We used the chat.nk function to estimate overdispersion in a 3-year study on elk that we are doing. We used the most global model and found no overdispersion for males during any year and overdispersion for females only during the last year of sampling. The first 2 years had chat values close to 1. My question is whether we should adjust all the SEs on D or just the last year? Which chat should we used for the QAICc calculation?
Thanks!
Thanks!
Murray Efford
May 17, 2023, 5:40:24 AM5/17/23
to jcl...@utk.edu, secr
Hi Joe
c-hat is estimated with error. David Fletcher advises against applying a correction unless the calculated c-hat lies outside the range expected by chance. In the coming release of secr that can easily be checked by simulation (new argument 'nsim' in chat.nk() triggers simulations). Release is a week or two off; otherwise get the development version from GitHub.
If some sessions have significant overdispersion and others don't then my preference is to base the decision on biological judgement: adjust if it makes sense because you can see a biological explanation for the difference. At least we are just talking about CI width - not like the bad old days where CAPTURE had people jumping between 'preferred' estimators, with different biases.
QAIC: a very good question, and one I've asked David to think about. Possibly c-hat pooled (averaged) across sessions, but what should be the weights? Similar sample size would argue for equal weights.
As an early adopter you may have to be patient as we sort out the wrinkles!
c-hat is estimated with error. David Fletcher advises against applying a correction unless the calculated c-hat lies outside the range expected by chance. In the coming release of secr that can easily be checked by simulation (new argument 'nsim' in chat.nk() triggers simulations). Release is a week or two off; otherwise get the development version from GitHub.
If some sessions have significant overdispersion and others don't then my preference is to base the decision on biological judgement: adjust if it makes sense because you can see a biological explanation for the difference. At least we are just talking about CI width - not like the bad old days where CAPTURE had people jumping between 'preferred' estimators, with different biases.
QAIC: a very good question, and one I've asked David to think about. Possibly c-hat pooled (averaged) across sessions, but what should be the weights? Similar sample size would argue for equal weights.
As an early adopter you may have to be patient as we sort out the wrinkles!
Murray
[apologies if you receive multiple copies of this - the software has been playing up for me]
We used the chat.nk function to estimate overdispersion in a 3-year study on elk that we are doing. We used the most global model and found no overdispersion for males during any year and overdispersion for females only during the last year of sampling. The first 2 years had chat values close to 1. My question is whether we should adjust all the SEs on D or just the last year? Which chat should we used for the QAICc calculation?
Thanks!
--
You received this message because you are subscribed to the Google Groups "secr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to secrgroup+...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/secrgroup/484d5349-e90b-4c03-886f-02e0875eda26n%40googlegroups.com.
jcl...@utk.edu
May 17, 2023, 8:31:09 AM5/17/23
to secr
Sound like you are on it Murray, as usual. The elk are gregarious animals of course but almost all the scats we collected were in and around relatively about 10 small fields in the area. We used a distance to field covariate to account for that in the model. My thought is that since the animals are grouped based on habitat patchiness, that the overdispersion caused by herding behavior is minimized. We happened to have a couple of transects that last year that went right through an active elk herd so we got lots of samples in year 3 at that site. I think that is accounting for a lot of the overdispersion we are getting that year.. My gut tells me that the CIs are not too narrow without adjusting with chat.
Reply all
Reply to author
Forward
0 new messages