Observer effects with occupancy estimation in MARK

[jluscie]

Is there a way to model survey-specific observer effects in the occupancy estimation algorithm in MARK?? If so, how is this coded in the '.inp' file?? Is there any documentation regarding this (I couldn't find anything)??

THANKS!!

Jason Luscier
University of Arkansas

[jlaake]

You haven't done so, you should spend some time reading the Cooch and White book because that will make it obvious that with MARK you can build such a model but MARK doesn' provide any specific models for you except some limited pre-defined models for things like group and time.

If each site was only visited by a single observer, then you can create groups of sites based on observer. Or if each occasion was done by a single observer then time effects can be used to model observer. However, if different observers were used for the same site at differing occasions, then you'll have to use a time-varying individual covariate which will get messy if you have many observers because if you have k observers then you'll need k-1 individual covariates for each occasion. So for example, if you have 5 observers and 10 occasions, you'll need 40 individual dummy variables (0/1) describing which observer was at each site on each occasion. Then you'll create and intercept (for one designated observer) and then you'll create k-1 columns for each of the remaining observers. In those columns, you'll want to enter the names of the variables for that observer-occasion. Using a smaller example, with 3 observers and 3 occasions with variable names obsij for observer i and occasion j, the DM for p would be

1 obs21 obs31
1 obs22 obs32
1 obs23 obs33

where observer 1 is the intercept. This type of example is described in the RMark documentation dicussion of occupancy models (Appendix C of Cooch and White).

--jeff