Temporary Removal

[brink2010]

I have encounter histories of different length due to temporarily removing individuals from the population for processing. For example, any animal I catch tonight will be processed tomorrow morning and returned AFTER sampling tomorrow night. Tomorrow night the individual will not be available for capture. Some of my individuals were removed for more than one night.

I read a few chapters of the Mark book on this website and found that you cannot use encounter histories that vary in length. Any suggestions on how I should proceed would be appreciated. Should I only use the individuals with the most common encounter history length (78 out of the 87 individuals)? Would it matter that they were out of the population during different surveys? Do I need to look into other means of calculating survivability and population estimates besides MARK?

Thank you!

[jlaake]

One possible solution is to keep the capture histories the same length by putting a 0 for the occasion in which the individual was unavailable. Then create a time varying covariate that has value 0 except that it has a value 1 for the occasion on which each individual was removed. The beta for that covariate should be large negative value because p=0. It should be identified as singular and can be removed from the parameter count. If the covariate was named rem1,rem2,...remk where k is number of occasions then in the column for that parameter use "remj" for the row that corresponds to p for occasion j. If you are trying to estimate abundance not sure how the models will perform but this will work for survival.

--jeff

[darryl]

You can also use a '.' in the encounter history at the corresponding occasion the indicate that no attempt was made to capture the animal that time, which achieves the same thing Jeff was suggesting, but without having to use a covariate.
Darryl

[jlaake]

Much simpler. Forgot about the "." notation.

[brink2010]

Thank you very much gentleman!

[Morten Frederiksen]

Or you can split each of these encounter histories into two dummy individuals, using the losses at capture coding (-1 in the count column) at time t for the first dummy individual, and reinserting the second at time t+1.

[brink2010]

Whenever I try to run a model on capture histories including "." I get an ERROR message that "Encounter history must consist of only these characters '01'." How do I get past this?

I looked at ch18 where an example used "." in the capture histories but it did not mention how to account for including values besides 0 or 1.

All suggestions are greatly appreciated!

[dhewitt]

A quick little search reveals that the dot "." notation in encounter histories is only allowed in certain model types. What model are you using?

viewtopic.php?f=21&t=1745&p=5178&hilit=dot+history+model#p5147

[brink2010]

I have a Robust design and am interested in survivorship and capture probabilities.

I currently have "." for nights I didn't survey and nights that certain individuals were temporarily unavailable.

[gwhite]

You have to use one of the Huggins data types with closed captures if you have dots in the encounter history. Dots will not work with the full likelihood data types because you would not know where dots should be included for animals never captured that are part of the full likelihood (but not Huggins).

Gary