We have a data on exercise (in hrs/week) and health outcomes for 10,000 people over age 60. We regress health on exercise and find a strong positive relationship with a high R-squared, low standard errors, etc. (i.e. good regression diagnostics). We conclude that exercise improves health. But this positive result may arise because healthier people are able to exercise more. a) The good regression diagnostics tell us that reverse causality is not a problem. b) Good regression diagnostics are necessary but not sufficient to infer from this data that exercise improves health. C) If causality runs in both directions the coefficient on the exercise coefficient is likely to overstate the true effect of exercise on health. d) b) and c). e) All of the above