bestglm: Best Subset GLM

The function bestglm selects the best subset of inputs for the glm family. The selection methods available include a variety of information criteria as well as cross-validation. Several examples are provided to show that this approach is sometimes more accurate than using the built-in R function step. In the Gaussian case the leaps-and-bounds algorithm in leaps is used provided that there are no factor variables with more than two levels. In the non-Gaussian glm case or when there are factor variables present with three or more levels, a simple exhaustive enumeration approach is used. This vignette also explains how the applications given in our article Xu and McLeod (2010) may easily be reproduced. A separate vignette is available to provide more details about the simulation results reported in Xu and McLeod (2010, Table 2) and to explain how the results may be reproduced.