Small sample bias in dynamic occupancy models

Abstract

Occupancy models may be used to estimate the probability that a randomly selected site in an area of interest is occupied by a species (ψ), given imperfect detection (p). This method can be extended, given multiple survey periods, to permit the estimation of seasonal probabilities of ψ, colonization (γ), persistence (φ), and extinction (1 − φ) in season t. We evaluated the sampling properties of estimators of these parameters using simulated data across a range of the parameters, differing levels of sites and visits, with a published dynamic occupancy model (Royle and Kery 2007). Bias depended largely on p and the number of visits, but also on the number of sites, ψt, γ, and 1 − φ. To decrease bias in all parameters to near zero, our results suggest that the number of required visits will depend on p, such that the probability of detection at an occupied site is near 0.9, and the required number of sites will be near 60 for ψt estimation and 120 or greater for γ and 1 − φ estimation. Published 2012. This article is a U.S. Government work and is in the public domain in the USA.

Effects of season on occupancy and implications for habitat modeling: the Pacific marten Martes caurina

“…Our generally high values for p made this less necessary (Royle 2006). However, future work should consider increasing the number of sites, since bias in estimating occupancy is least when the number of sites 60 (McKann et al 2013).…”

“…Martens typically have high probabilities of detection (Kirk andZielinski 2009, Slauson et al 2012). Our high estimate of detection probability provided some insurance against bias in each of the parameters of interest (p, ) (McKann et al 2013). Exceptions were the relatively low estimates of and p at Lassen and Sagehen.…”

Full Issue

“…Assuming that conditions identified by Mckann et al (2013) can be generalized to 2-species models, it is possible that the smaller number of sites in some study areas (RAI, NWC, and HUP) led to bias in parameter estimates. However, Mckann et al (2013) assumed a relatively short time series (5 yr) in their simulations, as compared with 19 yr for most territories in our analysis, and longer time series may compensate for fewer territories in some areas. Lastly, we defined Barred Owl occupancy in terms of one or more Barred Owls because the incidental detection data that we had available did not distinguish single owls from pairs (Yackulic et al 2012(Yackulic et al , 2014.…”

“…Despite these methodological differences, the results reported by Yackulic et al (2014) are similar to ours, suggesting that any heterogeneity that was unmodeled in our analysis may have been unimportant. The potential for small-sample bias has been explored in single-species, multiseason occupancy models (e.g., Mckann et al 2013), but has not been explored for 2-species, multiseason occupancy models. Assuming that conditions identified by Mckann et al (2013) can be generalized to 2-species models, it is possible that the smaller number of sites in some study areas (RAI, NWC, and HUP) led to bias in parameter estimates.…”

“…The potential for small-sample bias has been explored in single-species, multiseason occupancy models (e.g., Mckann et al 2013), but has not been explored for 2-species, multiseason occupancy models. Assuming that conditions identified by Mckann et al (2013) can be generalized to 2-species models, it is possible that the smaller number of sites in some study areas (RAI, NWC, and HUP) led to bias in parameter estimates. However, Mckann et al (2013) assumed a relatively short time series (5 yr) in their simulations, as compared with 19 yr for most territories in our analysis, and longer time series may compensate for fewer territories in some areas.…”