Ticket prices, diagnosed. Exercises in Chapter 18 considered weekly receipts from shows on Broadway in New York City. To simplify matters, we omitted a few weeks from those data. Here is the same regression with all the data present.

a) The two points plotted with x in the residuals vs. predicted plot are the two with the highest Cook’s Distances.
What does this information say about how those points might be affecting the analysis?
In fact, those points are reports published the last week of November and the first week of December in 2007—a time when the Stagehands Union Local One was on strike, closing down most Broadway shows. It seems that these are not representative weeks for business on Broadway.
Removing them results in the following regression.

b) What changes in these estimated models support the conclusion that those two weeks were influential points?
(Hint: The increase in adjusted isn’t one of the reasons.)

c) Which model would be the best to use in analyzing the business of Broadway shows—the one with all the data or the one with the two influential points removed? Explain.

Transcribed Image Text:

Dependent variable is: Receipts ($M) R-squared 93.3% R-squared (adjusted) = 93.1% s = 0.9589 with 92-4 = 88 degrees of freedom Sum of Squares df Mean Square F-ratio Source Regression 1130.431 3 376.811 88 0.919395 410 Residual 80.9067 Variable Coeff Intercept -22.1715 SE (Coeff) t-ratio P-value 2.221 -9.98 0.0001 Paid Attendance Number 0.087031 0.0046 19.0 0.0001 of Shows -0.024934 0.0338 -0.737 0.4632 Average Ticket Price 0.265756 0.0286 9.29 0.0001 Here are some diagnostic plots. Residuals 0 -2 -4 -6+ 10 15 20 25 25 Predicted