The claim that better fitting models lead to better decisions is intuitive, but can we make this claim explicit? Students learn that getting a higher R-square in their models typically means a tighter prediction interval. From the business side, does this improvement mean a better bottom line? This talk gives an example in which we can illustrate how improved precision translates into higher profits. Changes in R-square convert directly into reductions in costs in the classic news vendor problem - but not in the expected way. An interactive example reviews this demand problem and illustrates that profits come from being right on average more than being precise. |
Updated 02/21/2015