How to save slope/beta coefficients from one regression and use after second regression?

[Nouschka Snijders]

Hey all,

I have to compare a beta coefficient from one regression with interaction terms (factor-variables) to a beta coefficient from another regression with interaction terms. How can I save the beta coefficients from the first regression?

My do-file look as follows:
clear all
set more off
import excel "\\studfiles.campus.uvt.nl\files\home\home07\u 1246 790\Master\THESIS\Data\TOTAL REGRESSION DATA.xlsx", sheet("Sheet1") firstrow

fvset base none ETF Market

/* Option 1: Per ETF */
regress TE5 i.ETF##c.VOL i.ETF##c.SKEW i.ETF##c.KURT i.ETF##c.DIV i.ETF##c.PD
save "\\studfiles.campus.uvt.nl\files\home\home07\u 1246 790\Master\THESIS\Performance Measurement\voorbeeld.dta"
regress TE5 i.Market##c.VOL i.Market##c.SKEW i.Market##c.KURT i.Market##c.DIV i.Market##c.PD
append using "\\studfiles.campus.uvt.nl\files\home\home07\u 1246 790\Master\THESIS\Performance Measurement\voorbeeld.dta"
test i28.ETF#c.PD == i2.Market#c.PD

And the error is as follows:
. test i28.ETF#c.PD == i2.Market#c.PD
variable ETF not found
r(111);


Is there a way I can save my slope coefficient from the first regression (i.e. regress TE5 i.ETF##c.VOL i.ETF##c.SKEW i.ETF##c.KURT i.ETF##c.DIV i.ETF##c.PD)

Thank you in advance

[Clyde Schechter]

Run -help estimates- and read about -estimates store- and -estimates restore-. -test- will only work with currently posted estimates, so you have to restore the first estimates if you want to -test- something from there.

[Marcos Almeida]

Welcome to the Stata Forum/Statalist.

Please make sure the command and output are easily readable.

For this, you may use the CODE delimiters or install the SSC dataex, as recommended in the FAQ.

Thanks.

[Nouschka Snijders]

Marcos Almeida Thank you, I will!

Clyde Schechter Thank you. I have tried "estimates", but the problem is that -test- does not work with interaction terms in my case, but -test- does work without interaction terms. How can I solve this problem?

[Marcos Almeida]

Hello Nouschka,

I wonder whether you have tried to create an interaction variable, instead.

[Nouschka Snijders]

Marcos Almeida Can you explain in more detail what you mean by interaction variables? I have created interaction terms by multiplying a categorical variable by a continuous variable. But those interaction terms are referred to as virtual variables, which mean they act like variables, but do not exist in the dataset. Which is basically my whole problem, because I want those variable in the dataset... How can I solve that?

[Marcos Almeida]

Hello, Noushcka,

Please see the example below:

. sysuse auto
(1978 Automobile Data)

. regress mpg i.foreign##c.price

      Source |       SS           df       MS      Number of obs   =        74
-------------+----------------------------------   F(3, 70)        =     17.17
       Model |  1035.73441         3  345.244804   Prob > F        =    0.0000
    Residual |  1407.72505        70  20.1103578   R-squared       =    0.4239
-------------+----------------------------------   Adj R-squared   =    0.3992
       Total |  2443.45946        73  33.4720474   Root MSE        =    4.4845

---------------------------------------------------------------------------------
            mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
----------------+----------------------------------------------------------------
        foreign |
       Foreign  |   10.41949   2.914673     3.57   0.001     4.606356    16.23262
          price |  -.0007723   .0002028    -3.81   0.000    -.0011767   -.0003679
                |
foreign#c.price |
       Foreign  |  -.0008195   .0004248    -1.93   0.058    -.0016667    .0000276
                |
          _cons |   24.51661   1.379351    17.77   0.000     21.76558    27.26763
---------------------------------------------------------------------------------

. gen for_price = foreign*price

. regress mpg i.foreign c.price for_price

      Source |       SS           df       MS      Number of obs   =        74
-------------+----------------------------------   F(3, 70)        =     17.17
       Model |  1035.73441         3  345.244804   Prob > F        =    0.0000
    Residual |  1407.72505        70  20.1103578   R-squared       =    0.4239
-------------+----------------------------------   Adj R-squared   =    0.3992
       Total |  2443.45946        73  33.4720474   Root MSE        =    4.4845

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
     foreign |
    Foreign  |   10.41949   2.914673     3.57   0.001     4.606356    16.23262
       price |  -.0007723   .0002028    -3.81   0.000    -.0011767   -.0003679
   for_price |  -.0008195   .0004248    -1.93   0.058    -.0016667    .0000276
       _cons |   24.51661   1.379351    17.77   0.000     21.76558    27.26763
------------------------------------------------------------------------------

Hopefully that helps.

[Nouschka Snijders]

Marcos Almeida Yes, it worked!! Thank you very much!

[Marcos Almeida]

I'm glad it worked well. Please keep in mind that factor notation is preferable to handle interaction terms. With margins, for example, this is a crucial point IMHO, seldom do we need to create an interaction variable, and your case seems to be on of them.