Correction to “Doubly-robust dynamic treatment regimen estimation via weighted least squares,” by Michael P. Wallace and Erica E. M. Moodie; 71 (3), 636–644, September 2015

The authors of ‘Doubly-robust dynamic treatment regimen estimation via weighted least squares' would like to correct the following error in their paper:

Section 3.1: The simulations presented in the paper were generated with outcome Y_i ∼ N($e^{x_i}+x_i^3$ + a_i(1 + x_i), 1) and not Y_i ∼ N($e^{x_i}-X_i^3$ + A_i(1 + X_i), 1), as originally written.

Section 3.2: The simulations presented in the paper were generated with covariate X_2i ∼ N(0, 1) and not X_2i ∼ N(A_1i, 1), as originally written. Note that if X_2i were dependent on A_1i then consistent blip parameter estimators would still be obtained if $Y_i^{opt}$ was restricted to depend only on X_1i and not X_2i.

Note that neither of these alterations affect any of the theoretical claims or simulation results presented in the paper.