Presence‐Only Data and the EM Algorithm

Abstract

Summary In ecological modeling of the habitat of a species, it can be prohibitively expensive to determine species absence. Presence-only data consist of a sample of locations with observed presences and a separate group of locations sampled from the full landscape, with unknown presences. We propose an expectation–maximization algorithm to estimate the underlying presence–absence logistic model for presence-only data. This algorithm can be used with any off-the-shelf logistic model. For models with stepwise fitting procedures, such as boosted trees, the fitting process can be accelerated by interleaving expectation steps within the procedure. Preliminary analyses based on sampling from presence–absence records of fish in New Zealand rivers illustrate that this new procedure can reduce both deviance and the shrinkage of marginal effect estimates that occur in the naive model often used in practice. Finally, it is shown that the population prevalence of a species is only identifiable when there is some unrealistic constraint on the structure of the logistic model. In practice, it is strongly recommended that an estimate of population prevalence be provided.

One-Class Collaborative Filtering

“…This approach is similar to density estimation [4]. When unlabeled data are available, a strategy to solve one-class classification problems is to use EM-like algorithms to iteratively predict the negative examples and learn the classifier [28,17,26]. In [9], Denis show that function classes learnable under the statistical query model are also learnable from positive and unlabeled examples if each positive example is left unlabeled with a constant probability.…”

Predicting Spatial Distribution of Key Honeybee Pests in Kenya Using Remotely Sensed and Bioclimatic Variables: Key Honeybee Pests Distribution Models

“…Automated random sampling of pseudo-absence/background samples from a set of pixels within the boundaries of Kenya [41], where these species have not been detected [42] was used to maximize predictivity. Pseudoabsence of the probability of presence which would otherwise be confined to presence only [43]. Each species was assumed to have the same probability of being anywhere on the landscape, hence every pixel or environment on the landscape had the same probability of being tagged as "background" in geographic and environmental space.…”

Strengths and Weaknesses of Quantitative Climate Reconstructions Based on Late-Quaternary Biological Proxies

“…The modern distributional data used in such modelling should include not only presences but also absences (Phillips & Dudik 2008;Phillips et al 2009;Soberón & Nakamura 2009;Ward et al 2009;Lobo et al 2010, but see Phillips & Elith (2010) for using presence data only) if probabilities of occurrence are to be estimated (cf. Argáez et al 2005).…”