Attributing survey propensity scores to a general population

[David Latterman]

Hello. I'm conducting a survey where I have a primary dependent variable (would you do x or y), and a bunch of demographic variables. I need to create a propensity score (probability to do x or y) based on the dependent variable and a few of the independent demographic variables, like age, gender, etc. THEN, I need to take that score and apply it to another database in order to predict the behavior of the people in that database. The people in the second database never saw the survey; the matching is solely on the independent variables.

I believe PSMATCH or TEFFECTS won't work because I don't have a treated/untreated sample. I imagine I can do a logistic regression based on my data, but how do I "attach" that to the second database?

Thank you!
David

[Clyde Schechter]

Yes, you can do the logistic regression on your survey data and store the estimates (-help estimates store-). Load in the new data set, and then reactivate the estimates with -estimates restore-. Now run -predict-: it does out of sample predictions.

(Actually, you don't even need to store and restore the estimates if after your first logistic regression you don't do anything that clears the estimates from memory, like running another regression command, or -clear*-.)