Release GOF Error, Intervals, and Covariates

posts related to the RMark library, which may not be of general interest to users of 'classic' MARK

Release GOF Error, Intervals, and Covariates

Postby as4006 » Fri Mar 25, 2022 7:23 pm

Greetings Folks,

I am a new user to RMark and have a few questions... I am using a ten year data set (2011-2021) in an open CJS model for survival and capture probabilities of female saltmarsh sparrows and seaside sparrows. When I run the gof test I get an error. Code and error are as follows:

co_sp<-import.chdata("F_Species_Survival.txt", header=T)
co_sp.proc=process.data (co_sp,model="CJS",groups=c("Species", "Site", "Year", "Latitude"),begin.time=2011)
co_sp.ddl=make.design.data(co_sp.proc)
summary(co_sp)

ch Species Site Year Latitude
Length:3275 SALS:2532 OC : 368 2012 :517 39.5056 : 368
Class :character SESP: 743 MW : 318 2014 :486 39.5363 : 318
Mode :character CL : 270 2013 :484 43.0411 : 270
AT : 265 2015 :424 39.6973 : 265
HM : 253 2011 :357 41.2611 : 208
ID : 188 2021 :301 40.653045: 188
(Other):1613 (Other):706 (Other) :1658
# Goodness-of-Fit Test ----------------------------------------------------
release.gof(co_sp.proc)

Length of command exceeded MAXCHR/2 characters.
RELEAS ERROR TERMINATION
Error in (x3 + 4):length(out) : argument of length 0

Any suggestions of how to work through this error?

Second question... Currently, I have 9 intervals and they are formatted as "in between the breeding season." For example, interval 1 = captures between April15th-April 29th, interval 2 = captures between April 30th-May 13th, interval 3 = captures between May 14th-May 28th, etc.. My question is, is it ok to have my intervals formatted as "in between the breeding season" or will RMark treat these intervals as years? I ask because I am using year as a covariate.

Any information is greatly appreciated! Thanks!
as4006
 
Posts: 5
Joined: Fri Mar 25, 2022 5:27 pm

Re: Release GOF Error, Intervals, and Covariates

Postby jlaake » Sat Mar 26, 2022 11:10 am

Your capture history is too long for Release. But with 9*10=90 occasions, I wouldn't have thought that was the case so something else may be going wrong. Please send me the file you are using to jefflaake@gmail.com and I will run and see what the issue is. You can have intervals within the year but you must specify time.intervals or they are all assumed to be the same length of 1. I would hope that Release works with unequal time intervals but to be honest I'm not sure it does. Please read the workshop notes (see link when you type library(RMark) and Appendix C of Cooch and White if you haven't already done so.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Release GOF Error, Intervals, and Covariates

Postby jlaake » Sat Mar 26, 2022 11:14 am

Never mind. It isn't the length of the capture history. You must have too many groups. I suggest you use latitude as a numeric variable rather than factor and remove from the groups variable. Also remove Year from groups. You can always create a Year variable from time in the design data. That should reduce the number of groups to a reasonable number depending on the number of sites and species you have.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Release GOF Error, Intervals, and Covariates

Postby as4006 » Mon Mar 28, 2022 4:48 pm

Greetings Jeff,

Thank you for the prompt response, it is greatly appreciated. I took your advice and set latitude as a numeric variable and removed it and year from the groups. The gof test ran and the results are as follows:

RELEASE NORMAL TERMINATION
Chi.square df P
TEST2 80.2299 110 0.9852
TEST3 72.8279 156 1.0000
Total 153.0578 266 1.0000

Many of the Phi and p estimates were unidentifiable at many of my sites in the RELEASE OUT file. I have 29 sites which cover the breeding range of the saltmarsh sparrow (VA-ME) and also covers where SALS and SESPs breed sympatrically (VA-MA). Many of my southern sites, sites in VA and NJ, have female captures a lot earlier than the northern sites (CT, RI, MA, NH, and ME). This means, that occasion 1 and 2 often start with 0 individuals released at a fair amount of sites. Are these 0's on occasion 1 and 2 causing phi and p estimates to be unidentifiable? Or is it that some sites do not have a sufficient capture history? For example, one of my sites (SG) in Virginia has very minimal captures for SALS and SESP (4 individuals per species released thoughout all nine intervals).

Any information is greatly appreciated. Thank you for your time and consideration.
as4006
 
Posts: 5
Joined: Fri Mar 25, 2022 5:27 pm

Re: Release GOF Error, Intervals, and Covariates

Postby jlaake » Mon Mar 28, 2022 5:24 pm

I'm not surprised that you have many unidentifiable parameters. Release fits g*t for both Phi and p as the most general model. Your g is 2 species * 29 sites which is 58 and then if I understood correctly you have 90 occasions, so that is 5220 phi and 5220 p parameters for the most general model. All I can say is yikes. So unless you have say 50,000 observations (captures) you aren't likely to get anything reasonable from Release and even then the data would be sparse. Release is based on CJS model which conditions on the first capture and the "data" in the model are those occasions past the first release occasion, so leading 0's aren't a concern but Release may show them as unidentifiable - but I'm not sure. Either way it doesn't matter. But it sounds like they are migratory which is a bit of a problem if they are moving through your study area during the span of time that you are sampling within the year. Do they change sites? Not sure how you are using sites as groups if they do move because variables in groups are static.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Release GOF Error, Intervals, and Covariates

Postby as4006 » Mon Mar 28, 2022 5:57 pm

SALS and SESPs are migratory but they do not move between my sites very often, unless the sites are close to one another. There were only a handful of occasions where a bird had moved from one site to another. In this case, we kept the records with a relevant capture history and tossed the one record when it was captured at another one of my sites.

Is there a way to break this down where I can get some reasonable estimates from Release? Maybe break down capture histories by individual sites and run them all separately?

Ultimately, the capture probabilities are the most important parameters for my study.
as4006
 
Posts: 5
Joined: Fri Mar 25, 2022 5:27 pm

Re: Release GOF Error, Intervals, and Covariates

Postby jlaake » Mon Mar 28, 2022 7:25 pm

I'm not sure how much more I can help you. You may want to consider collapsing to annual occasions. Why are capture probabilities of interest? Usually they are a nuisance parameter except for abundance estimation.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Release GOF Error, Intervals, and Covariates

Postby as4006 » Mon Mar 28, 2022 8:27 pm

I am analyzing data from two different monitoring protocols and testing to see if there is a correlation among the two protocols. One is more of a standard population monitoring protocol (nest searching and monitoring, veg surveys and bird banding throughout the breeding season) and the other is a rapid assessment (similar survey methods but only two annual visits, one in early breeding season and one in late breeding season). Ultimately, the capture probability from the standard population monitoring protocol (the analysis we are discussing now) would be used to adjust capture probabilities from the rapid assessment protocol.

I have a meeting with my advisor tomorrow to discuss options. Maybe we will have to scale it back or, as you mentioned, collapse intervals into annual occasions. Thank you for your time and consideration, it is greatly appreciated!
as4006
 
Posts: 5
Joined: Fri Mar 25, 2022 5:27 pm


Return to RMark

Who is online

Users browsing this forum: No registered users and 16 guests

cron