14.10. For the Houses2 data file, fit the model to y = selling price using house size, whether the house is new, and their interaction.

(a) Show that the interaction term is highly significant.

(b) Show that observation 5 is highly influential in affecting the fit in (a).

(c) Show that the interaction effect is not significant when observation 5 is removed from the data set.

(d) Now fit the model in

(a) using a GLM assuming a gamma distribution for y. Note how the estimated interaction effect differs considerably from that in (a), and note that it is not significant. (Observation 5, highly influential for ordinary least squares, is not so influential in this analysis.

See also Exercise 13.22.)