The goal of this exercise is to construct a new statistic for NBA players, based on aspects of basketball that are not directly related to scoring, but which are important for winning games. Dataset 7.11 contains data on three NBA seasons, 2015-16 to 2017-18. For each team, in each season, several performance variables are recorded, along with the team’s average points scored per game minus its average points allowed per game (the variable DIFF). The performance variables are ORB, offensive rebounds, DRB, defensive rebounds, STL, steals, BLK, blocks, and TOV, turnovers; each of these is measured on a per-game basis for the team.
A. Find the linear function of ORB, DRB, STL, BLK, and TOV that has maximum correlation with DIFF. Note that, because each NBA team plays the same number of games, this linear function also maximizes the correlation if all variables are season totals (per game statistics were used for convenience).
B. Does the linear function make sense as a measure of performance for an NBA player? For instance, do the coefficients have the correct sign (positive or negative) and are the relative magnitudes of the coefficients appropriate?