Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.

Share this Page URL
Help

Nine: Examples > 9.6 A Multistate (Multistrata) Model - Pg. 245

EXAMPLES 245 Equation 9.10 is complicated and difficult to compute. Fortunately, both 9.10 and 9.10 (without the CV term) are computed by the MRAWIN soft- ware package. Running the fully time-dependent CJS model by sex on the dipper data (table 9.1), and requesting that F.cr.estim estimate vari- ances by inverting the Hessian of the maximization, produces the same parameter estimates as in Table 9.7, but slightly different standard error estimates. For male birds, the standard errors estimated by F.cr.estim for capture probabilities were 0.1980, 0.0000, 0.0822, 0.0612, and 0.0601 ^ ^ for p 2 through p 6 , respectively. For females, estimated standard errors were ^ ^ 0.2037, 0.1302, 0.0703, 0.0748, and 0.0567 for p 2 through p 6 , respec- tively. Applying (9.10) to this model resulted in the estimated standard ^ ^ ^ ^ errors se(N 2 ) = 20.23, se(N 3 ) = 7.77, se(N 4 ) = 6.30, se(N 5 ) = 6.39 and ^ ) = 5.86. Because the estimated CVs of capture probabilities were all se(N 6 <30%, it might be reasonable to apply (9.10) without the CV term. Doing ^ ^ so resulted in the estimated standard errors se(N 2 ) = 19.97, se(N 3 ) = 7.74, ^ ^ ^ ) = 6.29, se(N ) = 6.37, and se(N ) = 5.85. Even though it remains se(N 4 5 6 small, the largest difference between these two sets of variance estimates occurs at occasion 2 where CVs are highest.