James D. Nichols, James E. Hines, Darryl I. Mackenzie, Mark E. Seamans, and R. J. Gutiérrez. 2007. Occupancy estimation and modeling with multiple states and state uncertainty. Ecology 88:1395–1400.

Appendix A. A simulation study.

A simulation study was conducted to investigate the performance of the estimators of the model. The simulation was based on 100 sites visited five times each, with the following basic parameter values: ψ¹ = 0.684, ψ² = 0.521, p¹ = 0.681, p² = 0.771, δ = 0.438. In each simulation run, 1000 data sets were generated according to the input parameter values, and the resulting data were fit to a model with all detection and classification parameters constant over occasion and site, model ( ψ¹, ψ², p¹., p²., δ.).

FIG. A1. This figure plots the true value of each parameter of interest (θ, dashed line) against the mean of the estimates of this value ( , solid line). The difference between these two plots reflects bias. The five plots in the first column of Fig. A1 show each model parameter estimator (indicated on y axis) when ψ1 (x axis) is varied from 0.10 to 0.90 and all other model parameter values are held constant at their basic values listed above. The five plots in the remaining columns similarly show the performance of the estimator listed on the y axis as the parameter on the x axis is varied between 0.10 and 0.90.

FIG. A2. This figure shows the same basic structure as Fig. A1, except the plot is of the estimated standard deviation of the point estimates (our best estimate of the true , solid line) versus the average of the model based estimates of standard error () , dashed line). The plot thus addresses the issue of bias for the model-based estimates of . As with Fig. A1, the plots show the performance of the standard error estimates of the parameter on the y axis when the true value of the parameter of the x axis is varied from 0.10 to 0.90.

FIG. A3. This figure was designed to address a specific question about the importance of the classification parameter, δ, to the precision of the occupancy estimates. With all other parameter values held constant as listed above, we varied δ from 0.1 to 0.9 and then estimated all parameters under the model specified above. We plot as a function of the correct classification probability as a means of investigating the sensitivity of precision of other parameter estimates to this classification parameter.