Simulating in R, analyzing in RMark/MARK, estimation problem

[claudiapenaloza]

Hello All,

I simulated encounter histories for a Multistate Robust Design in R and have analyzed them in RMark and MARK but something is wrong.

The model has four states A, a, D and d ("a" is unobservable for "A", "D" and "d" are absorbing states). Transitions between "A" and "a", and detection for "A" are being estimated correctly but transitions between "A" and "D" or "d", and detection of "D" and "d" are not. In the later case, the transitions are estimated at half what they should be and the detection probability is being estimated twice as much as it should be.

Any idea why this could be happeing? (aside from the obvious "your simulation is incorrect", which is probably true but why then the coincidental 1/2 to 2x in paired transition/detection?)

Any help greatly appreciated,
claudia

[jlaake]

Claudia-

You may want to post this in a different sub-forum as it isn't really about RMark. Not sure that I have any insights other than that it makes sense that those parameters might tradeoff in that manner but beyond that not sure what to say.

regards --jeff

[claudiapenaloza]

After talking to Bill Kendall for a long time... we determined (non-rigorously) that these parameters are confounded. I am now going down another route.