Table 1:

Value function of the fitted ITRs with respect to the latent disorder severity (sum of Z₁; the lower, the better) and the predictive symptoms (sum of $\tilde{Y}$; the lower, the better) on the independent test set, mean (standard deviation), from 100 simulations under different training sample size N

N		Proposed	CausalForest	TARNet	Deep-O	Deep-Q	Deep-A
Value function for latent disorder severity (sum of Z1; the lower, the better)
200	Mean (sd)	1.80 (0.04)	1.85 (0.06)	1.93 (0.05)	1.85 (0.04)	1.83 (0.04)	2.00 (0.08)
500	Mean (sd)	1.77 (0.02)	1.80 (0.03)	1.91 (0.07)	1.80 (0.02)	1.80 (0.02)	1.98 (0.06)
1000	Mean (sd)	1.76 (0.01)	1.79 (0.02)	1.88 (0.09)	1.79 (0.01)	1.79 (0.01)	1.98 (0.04)
2000	Mean (sd)	1.74 (0.01)	1.79 (0.01)	1.87 (0.09)	1.79 (0.01)	1.79 (0.01)	1.98 (0.04)
Value function for predictive symptoms (sum of $\tilde{Y}$; the lower, the better)
200	Mean (sd)	12.74 (0.19)	12.91 (0.28)	13.30 (0.22)	12.92 (0.19)	12.84 (0.17)	13.62 (0.32)
500	Mean (sd)	12.59 (0.10)	12.70 (0.10)	13.22 (0.33)	12.72 (0.09)	12.71 (0.10)	13.53 (0.24)
1000	Mean (sd)	12.57 (0.09)	12.65 (0.06)	13.06 (0.40)	12.65 (0.06)	12.67 (0.06)	13.55 (0.17)
2000	Mean (sd)	12.52 (0.05)	12.62 (0.04)	13.01 (0.41)	12.64 (0.06)	12.67 (0.06)	13.54 (0.17)