Fig 2. Distributions of estimates the sizes of populations consisting of 1,000 (left) and 20,000 (right) members.

Estimation was based on 2–5 encounter event-histories generated from models ${\mathcal{M}}_h$, ${\mathcal{M}}_b$, ${\mathcal{M}}_t$, ${\mathcal{M}}_{bh}$, ${\mathcal{M}}_{ht}$, and ${\mathcal{M}}_{bht}$ for expected encounter probabilities of 0.025, 0.050, 0.100, 0.150 and 0.200 (right margins). LLM-AIC denotes selection of the single best LLM based on AIC, LLM-BMA denotes Bayesian model-averaging of loglinear models, and BLCM denotes nonparametric Bayesian latent-class model estimation. Estimates larger than 10⁷ are heaped at that value. Results for N = 10, 000 were similar to those for N = 20, 000.