Extracting regression coefficient values

[tim malz]

Hallo everyone:

I want to run regression for all firms in industry i to estimate specific regression coefficients for each industry. That's my very short example data set:

obs firm_id industry  y   x1    x2    x3
1    314      1      3.1  2.0  19.1   0.4
2    314      1      0.9  2.3  11.4   0.5
3    319      1      etc
4    401      2
5    408      2
6    409      3
7    409      3

I tried the following code:

sum industry
scalar a=r(min)
scalar b=r(max)

forvalues c=`=scalar(a)'/`=scalar(b)' {
qui reg y x1 x2 x3 if industry==`c'
capture noisily mat beta=e(b)
capture noisily svmat double beta, name(var_`c')
}

What I want to achieve are 4 variables wit the regression coefficient values (x1 x2 x3 intercept) by industry. However, my code generates 4 variables for every industry with the coefficient values only in the first observations; the data editor looks like:

obs firm_id industry  y   x1    x2    x3  var_11  var_12  var_13  var_14  var_21  var_22  var_23 var_24  var_31  var_32  var_33  var 34
1    314      1      3.1  2.0  19.1   0.4  0.78    0.13    0.78    0.64    0.69    0.20    0.51    0.64    0.83    0.03    0.78    0.44
2    314      1      0.9  2.3  11.4   0.5    .      .        .       .       .       .       .       .       .       .       .       .

I hope this explanation is understandable. Does anyone have an idea how to solve this problem, maybe with a combination of replace and svmat? I haven't found any working solution for this.

Thank you!
TM

[Nick Cox]

help statsby

[tim malz]

Sometimes it's that easy ... Thank you, Nick!