The first set (Table I) features a true model with three predictor variables, where each variable has the same influence on the response. The sample size is 20. SICf, greatly outperforms SIC: SICf correctly chooses the true model 98.1% of the time, compared to a 78.8% correct selection rate for SIC. Note that SICf, does not exhibit the same tendency as SIC to choose models which have too many predictor variables.