Automatic interaction detector (AID) and CHAID

Author: Paul Harris

Abstract:

In this technique we are interested how one key variable relates to other survey variables. The technique of AID is particularly useful when applied to survey tabulations. Let us suppose that we have interviewed 1,000 smokers and asked them. How many cigarettes do you smoke in an average day?' We would normally analyse the survey by such things as age, sex and region and interlaced categories of them. The idea would be to isolate age/sex/region groups which have very high and very low average smoking levels. In other words we are looking for differences in the key dependent variable smoking consumption. We are looking for categories of our survey which we may predict have high and low smoking levels.

  • PDF
  • This could also be of interest