rego: Stata module for decomposing goodness of fit according to Owen and Shapley values

Decomposition of the goodness of fit to (groups of) regressor variables can be a useful diagnostic tool to quickly assess “relative importanceâ€. Owen and Shapley values, two closely related solutional concepts in cooperative game theory, provide unique solutions to the decomposition exercise on the basis of a sound set of assumptions. At this stage, the new command rego implements decomposition of R-squared in OLS regression, based on the covariance matrix of the data for fast computation in the Mata environment. It also allows for bootstrapping the outcomes. Inclusion of other measures of fit and classes of models is planned for future extensions.