We now conclude our study of central tendency and regression this week. We have accomplished a lot in a short amount of time—learning how to calculate standard deviation, observing the regression to the mean, and identifying the correlation between an explanatory and response variable.
Andew Gelman and David Madigan touch on the ethical dimensions of Statistics in How is Ethics like Logistic Regression. Logistic regression is very similar to linear regression, except the response variable is categorical (yes/no), and not numerical. So, for example, a logistic regression model could take into account a variety of factors to determine whether we should implement a new medication, given the possible risks. In addition, a logistic regression could determine whether a homeowner is likely to default on a mortgage. This technique can be used to answer sensitive questions.
Often we need to make decisions in Statistics that might be considered unethical. Gelman and Madigan (2015) vividly claim “In general, though, the most informative ethics vignettes are those in which the call is not so close as to seem arbitrary, but not so obvious that the decision can be made without thought”. The best Statistical problems are those where the decision is not easy to make and a wrong decision carries a hefty cost.
Do you think regression should be used to answer sensitive problems, where a wrong decision incurs great risk? How strong of a correlation do you think is enough for us to feel a certain decision is viable? Can Statistics ever provide certainty for us in decision making?
Reference: Gelman, A., & Madigan, D. (2015). How is Ethics Like Logistic Regression. Chance, 28(2), 31-33. Retrieved May 8, 2019, from http://www.stat.columbia.edu/~gelman/research/published/ChanceEthics13.pdf