Fig. 4
From: Bayesian estimation of the discrete coefficient of determination
Bias, variance, and RMS for several CoD estimators versus sample size for d=2, under different values of the concentration parameter: Δ c /2=Δ p =Δ q =10 (high-variance priors), Δ c /2=Δ p =Δ q =50 (medium-variance priors), and Δ c /2=Δ p =Δ q =100 (low-variance priors). Plot key for base measures: dash-dot line (uniform prior), dashed line (mismatched prior), dotted line (poorly matched prior), and solid line (matched prior); see Table 1. Plot key for CoD estimators: MMSE (red), OBP (blue), resubstitution (gold), leave-one-out (purple), 0.632 bootstrap (green), 10-repeated twofold cross-validation (black). Results for the OBP CoD estimator are exact while others are approximated by Monte Carlo sampling method. The curves for the nonparametric CoD estimators and the flat prior Bayesian CoD estimators are repeated across the columns, for comparison with the results for the nonflat prior Bayesian CoD estimators