What is a typical question that can be investigated with TargetDiscovery?
TargetDiscovery is a tool for discovering the most important causal factors affecting an outcome. Given a list of potential causal factors (also referred to as target variables), TargetDiscovery carries out a separate analysis to estimate each factor’s effect on the mean of the outcome. The target variables are then ranked based on the strength of the evidence suggesting a non-zero effect (p-value).The entire discovery process can be performed separately for different subsets of the dataset, where such subsets can be defined by specifying one or more subgroup variables. In general, the causal link between a target variable and an outcome cannot be assessed by simply considering the association between the two variables. For instance, suppose the goal is to assess the causal link between carrying lighters and lung cancer. People who carry lighters are more likely to be smokers and will therefore show increased rates of lung cancer. A naive conclusion from this association between carryi
