1

I have a general question about propensity scores in R. I worked with SPSS Modeler in the past and I started use mainly R to build machine learning models instead. In SPSS Modeler, we used have an the option called "propensity score", which means that these scores shows the likelihood of an outcome.

So, is it possible in R to code for propensity score for each customer also? If so, could please direct me to it? Any other score also would work which gives the likelihood of an outcome such as confidence values (not confidence intervals) in R?

For example: What the likelihood of rain, not-rain, cloudy, and sunny in New York, Illinois, Georgia, Hawaii, and California in next week? Outcome: 5% Likely it is going to rain in NY, 60% likely it is going to be sunny in IL, 10% likely it is going to not rain in CA....etc.

Please the attached document from IBM SPSS Modeler on Page 39, this will give quick overview. I think they might have explain better. IBM_SPSS_PropensityScore Thank you!!!

Cettt
  • 11,460
  • 7
  • 35
  • 58
sharp
  • 2,140
  • 9
  • 43
  • 80
  • 1
    Propensity scores are just preliminary logistic models with treatment status as the outcome that are then used as stratification levels or matching criteria in other models where the outcome of interest is mortality or cost or something-else-that's-important. If you need an SPSS package to figure this out then maybe R is the wrong platform for you. – IRTFM Jan 30 '15 at 22:45

1 Answers1

1

Try package MatchIt for a variety of matching procedures. Help is on the author's website.

paulb20
  • 191
  • 3