Selects Columns with fewer missing values. This can be important for fitting models in which if any of the columns is missing, the whole row is dropped.