A grocery company sells all beer at the same price per six pack. Store managers can promote beer sales in three ways. They can advertise beer through weekly advertisements placed in local newspapers. They can also reduce beer prices either by a 15% promotional discount or a 25% deep discount. Management wants to know how these promotional options influence the quantity of six packs sold. You assemble a data set that includes weekly quantities of six packs sold (SPS) by company stores. Every record includes an attribute (AD) assigned a value of one if beer is advertised, zero otherwise, an attribute (PD) assigned a value of one if a promotional discount was offered during the week, zero otherwise and an attribute (DD) assigned a value of one if a deep discount was offered during the week, zero otherwise. You establish your margin of error as 5% and test the data model SPS = AD + PD + AD*PD + DD + AD*DD. Regression reports the following output: Parameter Estimate Test Statistic P-value Model n/a f = 12.47 0.0013 Constant 710 t = 3.41 0.0001 AD 38 t = 2.72 0.0108 PD 44 t = 2.24 0.0279 AD*PD 21 t = 1.98 0.0416 DD 67 t = 2.81 0.0062 AD*DD 33 t = 2.43 0.0137

What does the parameter estimate for AD predict about the number of six packs sold per week and why? Is this positive or negative and why?