Simulation results for Program Presence

[Kiel]

I am assessing the potential of different sampling regimes for a marsh bird monitoring pilot study for which species-specific information on both occupancy and detection probability are lacking. I am conducting these simulations in Program Presence and I am having difficulty with differentiating between the 'Simulation based estimate of standard error' and the 'Average estimate of the standard error'. Often these numbers are very similar, but in some instances they are not (i.e., when both detection and occupancy are <0.2). Which number should be used as the SE for the 'Average estimate of occupancy probability'?

Thanks, Kiel

[darryl]

Hi Kiel
For each simulated data set you can estimate psi and SE(psi). The 'simulation based estimate of SE' is the standard deviation of the 1000 (or however many simulations you run) estimated psi values and the 'average estimate of the SE' is the average of the 1000 (say) SE(psi) values. If there's no bias and there various asymptotic assumptions are holding, then Frequentist theory says these 2 values should agree. If they don't then either theres bias creeping in somewhere, or sample sizes are too small. Note that when p is getting small, you need to increase the number of repeat surveys otherwise the data set is too sparse, you start getting some bias, and things become unreliable.

Cheers
Darryl