Model comparison
Data Cutoff: July 10, 2026Results generated: 2026-07-14 00:41 UTC
| Model | LOO | LOO SE | Params | Est. N (95% CrI) | Dark figure (95% CrI) | Rank |
|---|---|---|---|---|---|---|
| M2 | -180.0 | 20.0 | 36 | 4,250.3 (2,685.8–7,335.1) | 3,074.3 (1,509.8–6,159.1) | 0 |
| M3 | -180.0 | 20.0 | 36 | 3,107.2 (1,881.9–5,298.9) | 1,931.2 (705.9–4,122.9) | 1 |
| M1 | -190.0 | 20.0 | 30 | 2,712.4 (2,454–2,990) | 1,536.4 (1,278–1,814) | 2 |
| M0 | -800.0 | 120.0 | 127 | 1,343.8 (1,319–1,370) | 167.8 (143–194) | 3 |
Selected model
Models M2, M3 are within ΔLOO < 2.0 of the best. Selected 'M2' as the simplest (fewest parameters).
LOO comparison
Model definitions & assumptions
M2
log(μ_c) = β₀ + Σ βₖ·Sₖ + Σᵢ<ⱼ βᵢⱼ·Sᵢ·Sⱼ + β_sex·sex + β_age·age_grp
Pairwise interactions (stratified). Extends M1 with all two-way source interaction terms (S₁·S₂, S₁·S₃, S₂·S₃). Captures positive or negative dependence between any pair of lists.M3
log(μ_c) = β₀ + Σ βₖ·Sₖ + Σᵢ<ⱼ βᵢⱼ·Sᵢ·Sⱼ + β₁₂₃·S₁·S₂·S₃ + β_sex·sex + β_age·age_grp
Three-way interactions (stratified). Adds the three-way interaction term (S₁·S₂·S₃) to M2, modelling the most complex dependence structure between all three lists.M1
log(μ_c) = β₀ + Σ βₖ·Sₖ + β_sex·sex + β_age·age_grp
Main effects (stratified). Adds individual source effects (S₁, S₂, S₃) as main effects alongside demographic covariates. Allows different base capture rates per source but assumes no pairwise dependence.M0
log(μ_c) = β₀ + β_sex·sex + β_age·age_grp
Independence model (pooled strata). Assumes all three lists capture individuals independently — no interactions between sources. Includes demographic covariates (sex, age group) only.