Skip to main content
Fig. 1 | Movement Ecology

Fig. 1

From: Incorporating periodic variability in hidden Markov models for animal movement

Fig. 1

Frequency of BIC-optimal numbers of states estimated, from 100 realizations of step lengths from an HMM with two latent states. In general misspecification increases the frequency of overfitting (i.e. selecting more states than the true value of n=2), whether temporal heterogeneity is neglected (red lines, bottom row) or observation error is not accounted for (right column). Fitting models may incorporate temporal heterogeneity in transition probabilities (blue lines) or exclude it (red lines). Subplots show variation in (left/right) whether error is added to the observed step lengths and (top/bottom) whether the simulation model incorporates temporal heterogeneity

Back to article page