in multiple regression, a technique by which the known variance in an outcome or response variable is separated out into the parts that can be uniquely attributed to each individual predictor variable and the parts that are common to any two or more variables.