Ecological Archives E085-072-A2

Juan Manuel Morales, Daniel T. Haydon, Jacqui Frair, Kent E. Holsinger, and John M. Fryxell. 2004. Extracting more out of relocation data: building movement models as mixtures of random walks. Ecology 85:2436–2445.

Appendix B. Hierarchical analysis of simulated movement data.

Here we present an example of a hierarchical analysis of movement trajectories applied to simulated data. We simulated movement of 10 individuals with fixed switching probabilities between a fast and a slow movement state. These 10 individuals chose all their movement parameters (scale and shape parameters for Weibull distributions used for step length, mean direction and mean cosine of turning angles for each movement state and switching rates) from the same "population level" distributions of parameter values. Simulated data was analyzed with a hierarchical version of the "Double Switch" model (see Supplement). Figure B1 shows that true individual parameter values where almos always well within the 95% credible interval produced by the model. The model also produced good estimates of population level parameters (Fig. B2), especially considering that only 10 individuals were used for these estimates.















   FIG. B1. True parameter values (solid dots) for 10 simulated movement paths and 95% credible intervals (thin lines) obtained with a hierarchical version of "Double switch" model. Parameters are scale of Weibull distribution for step length (a), the corresponding shape parameter (b), the mean direction of turning angles (), the mean cosine of turns () and switching probabilities (q).




   FIG. B2. Posterior distributions of population level parameters. The red lines are smoothed kernel densities from 5000 iterations of four Markov Chains. Arrows indicate true values.

[Back to E085-072]