What if your Y-variable is categorical?

• Administrivia:
  • Discussion of writting up project
  • Extra office hours by Nuria: 10 - 1 April 26 and 27th.

Examples

• medical: live/die (cured or not)
• psychological: clasification of mental health (but could be on a scale)
• in class: Fake-IDs (yes, or no -- maybe doesn't make sense)
• economics: buy/don't buy

How to deal with it

• Ignore it (code zeros and ones)
  • Advantage: just like regular regression
  • Advantage: graphs, leverage plots, usual intepretations
  • Disadvantage: Oops, negative probabilities (and great than 1)
  • Disadvantage: Inefficient, and incorrect SE's so incorrect t's
• Logistic solution (call it "nominal"--code with letters)
  • Advantage: moral high ground
  • disadvantage: bad graphics
• Logistic regression details
  • Curve instead of straight line
  • P(y) = exp(x)/(1 + exp(x))
  • P(y)/(1-P(y)) = exp(x) (odds ratio)
  • log(P(y)/(1-P(y))) = x (log-odds ratio == logit for short)

JMP example

• Challenger data

Last modified: Thu Apr 20 12:21:01 2000