Multi-session model

[Hugh Davies]

Hi there,

 

I am trying to estimate the density of native mammal species at four locations in northern Australia. These four locations are independent so I set up a multi-session capthist object to investigate which session specific covariate best explains any differences in density. My data for each of the four locations comes from two surveys (each of the two surveys consisted of four nights/occasions of live-trapping): the first survey was conducted in June 2019, the second survey was conducted in September 2019. It is likely the g0 and sigma vary between the two survey times as animals make larger movements later in the year as the dry season progresses.

 

I’m am a bit unsure on how best to model the potential difference in g0 and sigma between the two survey times and just wanted to double check that I’m on the right track..

 

I’ve attached my script and files, any advice would be much appreciated!

 

Cheers

Hugh

[Murray Efford]

Hi Hugh

Looking at your data, it seems you have concatenated occasions from the two surveys at each site, giving 8 occasions at each. The 'T' predictor relates to a linear trend across occasions on the link (logit) scale, which is probably not what you intend*. You can model a step change with a factor time covariate to fit one g0 (and/or sigma) for the first 4 occasions and another for the second 4 occasions. I guess you are happy to assume constant density across surveys. It might be more flexible and transparent to treat the data as 8 separate sessions (4 sites x two times); this does not acknowledge the continuity of the population at each site (same animals, probably same centres) that has an effect (probably minor) on the reported SE.

Murray

* although it actually gave a lower AIC than this model:

fit_CR_Fire2 <- secr.fit(CH, mask = mask, trace = TRUE, sessioncov = sesscov, verify = FALSE,
 model = list(D ~ Fire, g0 ~ tcov, sigma ~ tcov), timecov=c(1,1,1,1,2,2,2,2))