Absent or undetected? Effects of non-detection of species occurrence on wildlife-habitat models

摘要: Abstract Presence–absence data are used widely in analysis of wildlife–habitat relationships. Failure to detect a species’ presence an occupied habitat patch is common sampling problem when the population size small, individuals difficult sample, or effort limited. In this paper, influence non-detection occurrence on parameter estimates logistic regression models relationships was assessed using analytical and simulations. Two patterns were investigated: (1) random distribution among patches; (2) non-random which probability detecting species covaried with measurable variables. Our results showed that sensitive even low levels occupancy data. Both analytic simulation studies show yields bias estimation models. More importantly, direction affected by underlying pattern whether variable positively negatively related occupancy. For positive coefficient, yielded negative estimation, whereas linkage covariates produced bias. reversed, leading estimation. A release–recapture livetrapping study small mammals central Indiana, USA, illustrate magnitude typical field protocol varying intensity. Estimates error ranged from 0 23% for seven after 5 days sampling. We suggest many situations, between detection need be established correctly interpret