Skip to main content
. Author manuscript; available in PMC: 2017 Mar 1.
Published in final edited form as: Ann Appl Stat. 2016 Mar 25;10(1):32–53. doi: 10.1214/15-AOAS849

Table 2.

Estimated optimal values and error rates of sequential advantage selection (SAS), S-score and LASSO methods for the scenarios with the first choice of β (three important prescriptive variables: X1, X9 and X10). True value Q(gopt) is the mean response following the true optimal treatment regime gopt, while Q(ĝopt) is the mean response following the estimated optimal treatment regime ĝopt of the three methods. The mean response of each optimal treatment regime is obtained by the average outcome of Monte Carlo simulation using the true model with 10,000 replicates. Err.rate stands for error rate of the estimated optimal treatment regime, which is the proportion of individuals with misspecified optimal treatment following the estimated optimal treatment regime. The quantity in parenthesis is standard deviation of corresponding value.

Setting True Value SAS S-score LASSO
Q(gopt) Q(ĝopt) Err.rate Q(ĝopt) Err.rate Q(ĝopt) Err.rate
Randomized Study
Model I, ρ = 0.2 1.66 1.64(0.05) 0.072 1.41(0.13) 0.275 1.32(0.19) 0.436
Model I, ρ = 0.5 1.60 1.56(0.06) 0.103 1.38(0.08) 0.270 1.35(0.15) 0.449
Model I, ρ = 0.8 1.51 1.41(0.06) 0.182 1.40(0.03) 0.199 1.40(0.06) 0.473
Model II, ρ = 0.2 2.08 1.78(0.10) 0.299 1.79(0.11) 0.283 1.45(0.14) 0.460
Model II, ρ = 0.5 1.91 1.66(0.09) 0.285 1.62(0.08) 0.289 1.73(0.10) 0.466
Model II, ρ = 0.8 1.65 1.49(0.06) 0.243 1.50(0.04) 0.233 1.57(0.03) 0.484
Model III, ρ = 0.2 2.66 2.33(0.13) 0.312 2.30(0.13) 0.325 2.30(0.19) 0.432
Model III, ρ = 0.5 2.54 2.22(0.10) 0.327 2.18(0.08) 0.346 2.22(0.17) 0.442
Model III, ρ = 0.8 2.20 1.95(0.06) 0.314 1.98(0.05) 0.296 2.03(0.13) 0.459

Observational Study
Model I, ρ = 0.2 1.66 1.63(0.02) 0.100 1.43(0.12) 0.257 1.56(0.03) 0.334
Model I, ρ = 0.5 1.60 1.55(0.04) 0.125 1.43(0.07) 0.236 1.51(0.03) 0.319
Model I, ρ = 0.8 1.51 1.41(0.06) 0.190 1.41(0.02) 0.196 1.45(0.01) 0.311
Model II, ρ = 0.2 2.08 1.82(0.09) 0.286 1.83(0.11) 0.256 2.01(0.04) 0.314
Model II, ρ = 0.5 1.91 1.68(0.08) 0.278 1.66(0.08) 0.268 1.83(0.03) 0.310
Model II, ρ = 0.8 1.65 1.49(0.06) 0.251 1.51(0.03) 0.228 1.59(0.01) 0.309
Model III, ρ = 0.2 2.66 2.40(0.08) 0.289 2.35(0.13) 0.289 2.57(0.04) 0.320
Model III, ρ = 0.5 2.54 2.27(0.09) 0.303 2.23(0.09) 0.310 2.44(0.03) 0.318
Model III, ρ = 0.8 2.20 1.98(0.07) 0.301 2.01(0.06) 0.273 2.13(0.01) 0.312