Fig. 4
From: A comparison of statistical methods for deriving occupancy estimates from machine learning outputs
(a) Influence of data verification method on occupancy estimate. Verification methods include: no verification, 10 randomly selected files, and top-ten files with the highest classifier score. (b) Influence of temporal subsampling regime on occupancy estimate. Intervals used for subsampling were 0 (no subsampling), 10 min, and 30 min. Subsampling methods were either ‘random’, which involved selecting the first recording from each interval, or ‘max’ which involved selecting the file with the highest classifier score within each interval. Results were removed for variations that did not converge (e.g. several of the Kéry models). Decision thresholds of 0.01 and 0.77 were used for (a) and (b) respectively. Bars denote 95% confidence intervals and credible intervals for maximum likelihood and Bayesian approaches respectively.
