In backward elimination (also referred to as backward deletion), the least significant variables are methodically removed until only the most important ones are left.
Specifically, this pertains to a type of stepwise regression which is a statistical method of building a regression model (an analysis to evaluate relationship between variables) from a group of predictor variables (sometimes called independent variables). For instance, a researcher has a set of demographic data from a survey completed by 1000 college freshmen. The predictor or independent variables are age, sex, height, religion, major course, minor course, ethnicity, and blood type. The researcher would then like to use the data in scouting for female basketball players. Through backward elimination, the unimportant variables such as religion, ethnicity, and blood type are removed one by one. As a result, the most significant predictors, which are sex (females), height (at least 5’8” tall), and course (majoring or minoring in physical education), for forming the team are left.