Can TargetDiscovery be used with data from randomized controlled trials?
Yes, if the treatment of interest is dichotomous (e.g. new drug versus placebo) or ordered categorical (e.g. different drug doses). You would specify the treatment as the target variable and pre-treatment (baseline) covariates as confounders. This kind of analysis will provide an ANCOVA estimate of the treatment effect that is always valid since the bias reduction step in the targeted maximum likelihood estimator always succeeds if the treatment is randomized.
