=+c. Fitting the model with predictors x1 and x2 gave SSResid 5 27,454, whereas fitting with x1, x2, and x3 5 x1x2 resulted in SSResid 5 20519.

Using  5 .01, can we conclude that the x1x2 term adds useful information to a ‘reduced’

model containing only x1 and x2? Note: when g 5 1, the resulting F test gives the same conclusion as the t-test for whether a single variable (here, x1x2) contributes useful information to a model.