Huggins closed capture model

[hlegra1]

I want to estimate detection probability (capture probability) with the removal model using Huggins closed capture model in MARK. My input file includes an encounter history (3 occasions), 24 groups (representative of 24 species) and 4 individual covariates (all categorical dummy variables), totaling 7,728 individual records. I have set the c parameter to 0 for all intervals, and want to test 10 models.

I get the following error code when attempting to run models with and without individual covariates: “Error: not enough individual covariates specified in the encounter history. Exit code A30.” I am not sure if this is due to the way I’ve set up my input file, or to the selections I’ve made in the “Setup Numerical Estimation Run” window. Since my covariates are dummy variables (0,1), I unchecked “standardize individual covariates.” Under “Real Par. Estimates from Individual Covariates,” I am not sure whether I should check First, Mean or User Specified covariate values. None seem appropriate, though I've tried them all and still get the error.

Thanks very much for any input on this matter.

Holly

[bmitchel]

Holly -

This is almost certainly a problem with your input file. The way you've described it, your file should have 29 space-separated fields, followed by a semicolon. So a 111 history for species 2 with covariates of 1 0 1 and 0 would be:

111 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0;

Regarding "First, Mean, or User Specified", probably the best option is user specified, in which you would select a set of dummy covariate values (i.e. one of the 16 possible combinations) when you run the model. The "mean" option does not make much sense with dummy variables. Then the "real parameter" estimates in the output will correspond to predicted results for that set of covariate values.

Brian R. Mitchell

[hlegra1]

Thanks Brian,

My input file is as you described, except that the encounter history for each record is either 100, 010, 001. I will try again with the suggested user specified options. By choosing to use the user specified option, it sounds like I will need to test each of the 16 possible combinations of individual covariates for each of the 10 models. Have I interpreted this correctly?

Thanks very much.

Holly