Using Two-Stage Least-Squares Regression to Model Cross-Sales

An order-by-mail company has a book club and a CD club. Each month, they make special offers available to club members. The company wants to create a model for the month's total special offer purchases based on total book purchases, CD purchases, and the type of offer given to club members. Two-Stage Least-Squares Regression is appropriate to this situation because money spent on special offers is money not spent on books or CD's; thus, there is a feedback loop between the response and these two predictors.

Ninety-nine months of sales information is collected in cross_sell.sav. See the topic Sample Files for more information. The file also includes a variable, Special offer, displaying each month's special offer, which has also been recoded into two indicator variables, Appliance offer and Checks offer, that can be used as predictors in the regression procedures. Lastly, the monthly discounts (and log-discounts) offered to club members are also listed.

Next