a technique used in
regression analysis in which the goal is to forecast an outcome or response variable according to a subset of predictor variables narrowed down from a large initial set. In background elimination, all available predictors are included originally and then examined one at a time, with any predictors that do not contribute in a statistically meaningful manner systematically dropped until a predetermined criterion is reached. Also called
backward deletion;
backward selection;
stepdown selection. See also
stepwise regression.