Consider a line of regression that fits data points showing crime rate and population for all U.S. cities with populations between 1 million and 10 million people.
For which U.S. cities would this line of regression not provide very good predictions of crime rate?
What advice would you provide to a sociologist from China who would like to use this line of regression to predict the crime rate in Chinese cities?
What are some possible confounding variables that might covary with both crime rate and city population?
What would you say to someone who was convinced that high birth rates cause an increase in crime approximately 20 years later. Especially if he insists that this particular line of regression proves it (that is, assuming it has a positive slope)?