This concept can be extended to estimate the contribution of a single variable to a statistical indicator of goodness-of-fit of the adopted model, such as the R-squared (Huettner & Sunder, 2012; Shorrocks, 2013). More formally, given a full regression model with K explanatory variables (x1, x2, ..., xK), computing the contribution of each variable requires the estimation of all possible sub-models (considering all K! permutations of regressors):