Abstract
This paper is concerned with the problem of selection of variable in multiple regression with flattener of Stein-type. A matrix of multiple regression with flattener is given. A relation between Allen's prediction sum of squares (PSS) in ordinary regression and squared deviation discrepancy (SDD) in multiple regression with flattener is given. Allen's PSS may be still used as an aid in selection of predictor variables in multiple regression with flattener. A criterion proposed here is proportional to PSS within a given data.
