.rego

Purpose

rego is a Stata module that decomposes R² (share of explained variance) of an OLS model into contributions of (groups of) regressor variables with the help of Shapley or Owen values. By using “groups” of variables that belong to the same category — such as the variables that belong to a polynomial in age — computational effort is lower than in the “classical” Shapley decomposition without groupings. rego has an implemented option to bootstrap the decomposition results in order to obtain percentile confidence intervals.

Example

In the example below, rego is used to decompose the R² of a wage regression. The backslash (\) symbol delimits groups of regressor variables. The option (detail) requests the computation of within-group decomposition. This would take a considerable amount of time if the number of variables in a group is large; the example below required 0.25 seconds.

Download

rego is published under the terms and conditions of the GNU General Public License 3.¹ The program is still in development, and no warranty is provided regarding the soundness of results.

To install the current version for Stata 9 or higher, execute the following in the command window:

. net from http://www.marco-sunder.de/stata/
. net install rego

Alternatively, download this zip file and place its contents either into your working directory or into a different folder that you specify with the adopath command.

References

rego is written and maintained by Frank Huettner and Marco Sunder, University of Leipzig.

Huettner, F.; Sunder, M. (2012). “Axiomatic arguments for decomposing goodness of fit according to Shapley and Owen values.” Electronic Journal of Statistics, 6, 1239–1250. doi:10.1214/12-EJS710

@article{HueSun2012EJS,
    author  = {Frank Huettner and Marco Sunder},
    title   = {Axiomatic arguments for decomposing goodness of fit
               according to Shapley and Owen values},
    journal = {Electronic Journal of Statistics},
    volume  = {6},
    pages   = {1239--1250},
    year    = {2012},
    doi     = {10.1214/12-EJS710}
}