KNN post estimation: probabilities of classification

[Noor Sethi]

Hi statalisters,

I conducted an out-of-sample KNN prediction that predicts poverty status based on some variable: x (code pasted below).

* generate a simple dataset
  clear all
  set seed 2876
  set obs 20
  
  gen x = round(runiform(0,100),1)
  gen y = round(runiform(0,100),1)
  gen r = round(runiform(0,1),1)

* create a binary categorical income var
  gen poor = 1 if y < 20
  replace poor = 0 if y > 20
  label define poor_lab 0 "non-poor" 1 "poor"
  label values poor poor_lab
  
* run knn
  discrim knn x if r == 0, group(poor) k(3)
  estat list if r == 1

The post estimation command -estat list- gives the probability of classifying each observation as "poor" or "non-poor". I want to create a variable that is equal to these probabilities. That is, running this code, I want a variable called prob_nonpoor equal to 1 in observation 2, 1 in observation 3, .20 in observation 4, and so on... It doesn't seem that stata stores this information though.

I'd appreciate any help, thank you!