Logistic regression in SAS

---

Here we'll analyze the basketball data using logistic regression. The b-ball data contains data for 1192 shots by Sam Cassell in the NBA 2004-2005 season. The outcome variable is "make" which is 1 if the shot is made and 0 if the shot is missed. There are several predictors including:

• the shot's distance ("distance")
• score differential at the time of the shot ("scorediff")
• whether teammates Kevin Garnett ("kg") and Latrell Sprewell ("spree") are in the game
• whether the game is played at home ("home"),
• the opponent's field goal percentage allowed ("fgpA")
• the opponent's winning percentage ("totW")
• time remaining in the game ("time")
• days off between games ("daysrest")

To perform logistic regression using pull-down menus:

• load the data using import wizard
• open the analyst window and load the bball data
• click statistics->regression->logistic
• Define "make" as the dependent variable and enter "1" in the "Model Pr{ }" field to indicate that we want to model the probability of making a shot, not the probability of missing a shot.
• Add the continuous predictors (distance, scorediff, fgpA, totW, time, and daysrest) to the "quantitative" list.
• Add the categorical predictors (kg, spree, home) to the "class" list.
• Click "OK".

Or, you can use this sas script in the editor window ("descend" tells SAS to model the probability of a make, not the probability of a miss):

proc logistic data = bball descend;
class kg spree home;
model make = distance scorediff kg spree home fgpA totW time daysrest;
run;