Researchers are on the constant hunt to identify causal relationships. The term treatment effect denotes the average causal effect of a binary variable on a defined outcome. The term ‘treatment’ originated from the medical literature in which a group of observation was “treated” (yes for the binary variable) versus another group that was not treated (no for the binary variable). This term is now widely used across different disciplines often including outcome variables of interest for policy makers.
Stata already includes an extensive set of commands to estimate treatment effects. Stata 14 goes a step further and adds a new command stteffects which, like the existing teffects allows the users to estimate average treatment effects (ATEs), average treatment effects on the treated (ATETs), and potential-outcome means (POMs) but also allows users to model a combination of the outcome, treatment assignment and censoring. Further, stteffects also offers the options of estimating treatment effects by inverse probability weighting (IPW) through stteffects ipw, different regression-adjustment methods through stteffects ra through stteffects wra and stteffects ipwra allows a choice between two doubly robust estimators.
Stata 14 allows users to deal with endogenous treatments where the treatment assignment is correlated with the outcome through the new command eteffects which estimates ATEs, ATETs and POMs for continuous, count and binary outcomes.
To see a video of the new treatment effects commands, and other new features of Stata 14 please visit this page. For Timberlake training courses in Stata 14 please visit this page.