Covariates in mult-session models

[Tim]

Hi there,

I'm trying to gain a better understanding on how to appropriately incorporate covariates (e.g. sex) into my model, particularly for multi-session data.

For context, I have three sessions (i.e. years) of data. There is a male bias in my capture data, as well as differences in sigma between sexes (males larger home range) which I want to account for.

I've done basic modelling for each year separately, using the behavioral models on g(0). The 'bk' (site-specific behavioural response) model is consistently the highest rank. I'm now trying to add all years into the one model (as sessions) and also include sex as a covariate. The first year of data had fewer captures (particularly females), therefore the estimates of sigma were too unreliable to split between sex. However, the population is increasing so there is now much more recapture data to estimate sigma more accurately for each sex.

In this instance, it seems most appropriate for the model to include 'session' on density (D), 'bk' on g(0), and 'sex' on sigma. I just want to check I'm on the right track here, or if there's suggestions for what I should be doing instead.

Here's example code and output from secr.fit: 
 secr.fit(BB,  hcov = "V6", model = list (D ~ session, g0 ~ bk, sigma ~ h2), mask = mask, trace = FALSE, biasLimit = NA)

Fitted (real) parameters evaluated at base levels of covariates session = 2023, h2 = Female, bk = 0 link estimate SE.estimate lcl ucl D log 3.733728e-02 0.009802606 2.251023e-02 0.06193061 g0 logit 8.406281e-03 0.001422738 6.030572e-03 0.01170686 sigma log 4.245069e+02 67.754685050 3.110855e+02 579.28173467 pmix logit 1.824255e-01 0.000000000 1.824255e-01 0.18242552 session = 2023, h2 = Male, bk = 0 link estimate SE.estimate lcl ucl D log 3.733728e-02 0.009802606 2.251023e-02 6.193061e-02 g0 logit 8.406281e-03 0.001422738 6.030572e-03 1.170686e-02 sigma log 1.150656e+03 79.894976703 1.004417e+03 1.318187e+03 pmix logit 8.175745e-01 0.000000000 8.175745e-01 8.175745e-01 session = 2024, h2 = Female, bk = 0 link estimate SE.estimate lcl ucl D log 4.480090e-02 0.009832547 2.928631e-02 0.06853446 g0 logit 8.406281e-03 0.001422738 6.030572e-03 0.01170686 sigma log 4.245069e+02 67.754685050 3.110855e+02 579.28173467 pmix logit 1.824255e-01 0.000000000 1.824255e-01 0.18242552 session = 2024, h2 = Male, bk = 0 link estimate SE.estimate lcl ucl D log 4.480090e-02 0.009832547 2.928631e-02 6.853446e-02 g0 logit 8.406281e-03 0.001422738 6.030572e-03 1.170686e-02 sigma log 1.150656e+03 79.894976703 1.004417e+03 1.318187e+03 pmix logit 8.175745e-01 0.000000000 8.175745e-01 8.175745e-01 session = 2025, h2 = Female, bk = 0 link estimate SE.estimate lcl ucl D log 7.959835e-02 0.017254077 5.230075e-02 0.12114353 g0 logit 8.406281e-03 0.001422738 6.030572e-03 0.01170686 sigma log 4.245069e+02 67.754685050 3.110855e+02 579.28173467 pmix logit 1.824255e-01 0.000000000 1.824255e-01 0.18242552 session = 2025, h2 = Male, bk = 0 link estimate SE.estimate lcl ucl D log 7.959835e-02 0.017254077 5.230075e-02 1.211435e-01 g0 logit 8.406281e-03 0.001422738 6.030572e-03 1.170686e-02 sigma log 1.150656e+03 79.894976703 1.004417e+03 1.318187e+03 pmix logit 8.175745e-01 0.000000000 8.175745e-01 8.175745e-01

Appreciate any assistance,

Cheers

[Murray Efford]

I'm not seeing any problem here. There are always other models you could try (e.g. sex effect on g0 on addition to sigma).