Using Regression When Y = # Wins

#1
Hello,

This is my first post, and so I apologize if this is a dumb question. I'm considering using multiple linear regression to predict Y = # wins for a basement team based on several independent variables (both quantitative and dummy X variables).

From reading this forum, I know that logit and/or probit is the preferred approach to use if Y is expressed as a winning PERCENTAGE (i.e. #wins/total #games). However, since I plan to express Y as # wins instead of percent wins, should I still use logit or probit analysis? It would seem that I should since Y is still "nominal" (categorical) data, as opposed to a continuous measured variable.

Thank you!
Gary
 
#2
Hi Gary,

Why don't you use a Poisson regression or a negative binomial regression ? This is the best way to deal with data in the form of a count or rate.

Etienne