Treatment analysis trains a propensity model to predict the treatment variable. Its purpose is to help assess whether some assumptions about the treatment variable are violated: Treatment randomization (the treatment variable is perfectly randomized) and Positivity (any subset of the dataset has a non-zero probability of being assigned the treatment or the control).

Enabling treatment analysis will provide:

• additional insights dedicated to the treatment variable in the Result screen of the Analysis
• ability to predict the treatment probability of each new record with scoring and evaluation recipes and API scoring
• ability to use inverse propensity weighting method to compute metrics
• ability to use propensity-based diagnostics (treatment and propensity checks)