Traditional forward variable selection methods based on the multiple regression is known to be inefficient and not robust when distributions of observations are unknown, skewed, or heavy-tailed. So, we suggest a statistic using relative signs or direc...
Traditional forward variable selection methods based on the multiple regression is known to be inefficient and not robust when distributions of observations are unknown, skewed, or heavy-tailed. So, we suggest a statistic using relative signs or directions to substitute for the critical measure in the traditional forward variable selection method.