Regression with Stored Coefficients

[mehlika oz]

Hello I have a regression of treatment on some individual characteristics in data set 1:

reg treatment1 female_1 age_1.... etc

I would like to use stored coefficients of this regression:

_b[female_1], _beta[age_1] to be able to calculate the probability of being treated in another data set by using the individual characteristics in data set 2:

basically it would be like:

treatment2=_b[constant]_b[female_1]*female_2+ _beta[age_1]*age_2+....

How to calculate this outcome in an estimation logic? i.e. as if I am calculating inverse probability?

Thanks in advance

[Ken Chui]

One way could be predict. See point 9 on page 1, and then pages 5-7 in https://www.stata.com/manuals/rpredict.pdf.

Here is a short example:

sysuse auto, clear
keep if _n <= 37 // Keep only the first half (aka, pretend to be 1st data set)
regress mpg price trunk // Run the regression

sysuse auto, clear
keep if _n >= 38 // Use only the second half of the data (aka, 2nd data set)
predict yhat // Apply the old regression to compute y hat

* Double check:
generate yhat2 = 31.50203 - 0.0004055*price - 0.6443269*trunk

Results:

     +---------------------+
     |     yhat      yhat2 |
     |---------------------|
  1. | 16.63252    16.6326 |
  2. |   20.786   20.78606 |
  3. | 23.35763   23.35769 |
  4. | 16.34288   16.34303 |
  5. | 22.53001   22.53008 |
     |---------------------|
  6. | 22.62003    22.6201 |
  7. | 18.73096   18.73102 |
  8. | 23.71725   23.71734 |
  9. | 19.54641   19.54647 |
 10. | 16.26432    16.2644 |
     +---------------------+

[Carlo Lazzaro]

Mehlika:
welcome to this forum.
-statsby- is another option (warning: use -statsby- on a copy of your original dataset):

. use "C:\Program Files\Stata17\ado\base\a\auto.dta"
(1978 automobile data)

. help statsby

. statsby _b, by(foreign) verbose nodots: regress price weight length mpg

      Command: regress price weight length mpg
    _b_weight: _b[weight]
    _b_length: _b[length]
       _b_mpg: _b[mpg]
      _b_cons: _b[_cons]
           By: foreign


. list

     +-------------------------------------------------------+
     |  foreign   _b_wei~t   _b_length     _b_mpg    _b_cons |
     |-------------------------------------------------------|
  1. | Domestic   6.767233   -109.9518   142.7663   2359.475 |
  2. |  Foreign   4.784841    13.39052   -18.4072   -6497.49 |
     +-------------------------------------------------------+

.

[mehlika oz]

Thank you both Ken and Carlo