Storing names of interaction terms in logistic regression to be used with testparm in a loop

[Caterina Felici]

Hi,
I am trying to run testparm for multiple interaction coefficients (for a categorical variable with multiple levels) after logistic regression in a loop. After testparm I need to put the names of the interaction coefficients generated but I am not finding a way of accessing them through postestimation commands; from the Stata documentation they do not seem to be stored. Do you know if there is a way of accessing the names of the interaction coefficients generated so I can easily use them in a loop without copying and paste each of them?

Best

[Mike Lacy]

The return list after -logit- includes r(table), from whose column names you can obtain a list of all the variables in the model with an extended macro function. The following might start you in the right direction:

logit .....
local varnames: colfullnames r(table)
foreach v of local varnames {
  di "`v'"
}

[Ken Chui]

You can find out the names with an option called coefleg. Here is an example:

sysuse nlsw88, clear
logit married age i.race##i.collgrad, coefleg

Results:

-------------------------------------------------------------------------------------
            married | Coefficient  Legend
--------------------+----------------------------------------------------------------
                age |    -.02144  _b[age]
                    |
               race |
             Black  |  -1.029574  _b[2.race]
             Other  |  -.0159885  _b[3.race]
                    |
           collgrad |
      College grad  |  -.1069247  _b[1.collgrad]
                    |
      race#collgrad |
Black#College grad  |   .1709965  _b[2.race#1.collgrad]
Other#College grad  |   -.060614  _b[3.race#1.collgrad]
                    |
              _cons |   1.730005  _b[_cons]
-------------------------------------------------------------------------------------

And if you just mean the name of the whole set of interaction, then it's just the row title (race#collgrad), as:

testparm race#collgrad

[Jeff Pitblado (StataCorp)]

testparm uses _ms_extract_varlist to translate a
varlist specification to matrix stripe elements in the current
estimation results. While _ms_extract_varlist is undocumented,
it does have a help file.

Here is an example of this command in action.

. sysuse auto
(1978 automobile data)

. regress mpg turn trunk for#rep
note: 1.foreign#1b.rep78 identifies no observations in the sample.
note: 1.foreign#2.rep78 identifies no observations in the sample.

      Source |       SS           df       MS      Number of obs   =        69
-------------+----------------------------------   F(9, 59)        =     10.99
       Model |  1465.91929         9  162.879921   Prob > F        =    0.0000
    Residual |  874.283608        59  14.8183662   R-squared       =    0.6264
-------------+----------------------------------   Adj R-squared   =    0.5694
       Total |   2340.2029        68  34.4147485   Root MSE        =    3.8495

-------------------------------------------------------------------------------
          mpg | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
--------------+----------------------------------------------------------------
         turn |  -.8183495   .1731554    -4.73   0.000    -1.164833   -.4718663
        trunk |    -.25088   .1475067    -1.70   0.094    -.5460402    .0442802
              |
foreign#rep78 |
  Domestic#2  |    1.60522   3.145736     0.51   0.612    -4.689383    7.899823
  Domestic#3  |   .3552652   2.991137     0.12   0.906    -5.629985    6.340516
  Domestic#4  |   .3116473   3.215529     0.10   0.923    -6.122612    6.745906
  Domestic#5  |   7.159133   3.963637     1.81   0.076    -.7720857    15.09035
   Foreign#1  |          0  (empty)
   Foreign#2  |          0  (empty)
   Foreign#3  |  -1.342274   3.762394    -0.36   0.723    -8.870806    6.186259
   Foreign#4  |  -.5612612    3.23632    -0.17   0.863    -7.037122      5.9146
   Foreign#5  |    1.72808   3.261867     0.53   0.598    -4.798901     8.25506
              |
        _cons |   56.68481   7.139194     7.94   0.000     42.39931     70.9703
-------------------------------------------------------------------------------

. testparm for#rep

 ( 1)  0b.foreign#2.rep78 = 0
 ( 2)  0b.foreign#3.rep78 = 0
 ( 3)  0b.foreign#4.rep78 = 0
 ( 4)  0b.foreign#5.rep78 = 0
 ( 5)  1.foreign#3.rep78 = 0
 ( 6)  1.foreign#4.rep78 = 0
 ( 7)  1.foreign#5.rep78 = 0

       F(  7,    59) =    1.26
            Prob > F =    0.2872

. _ms_extract_varlist for#rep

. di "{p}" r(varlist) "{p_end}"
0.foreign#1.rep78 0.foreign#2.rep78 0.foreign#3.rep78 0.foreign#4.rep78 0.foreign#5.rep78
1.foreign#1.rep78 1.foreign#2.rep78 1.foreign#3.rep78 1.foreign#4.rep78 1.foreign#5.rep78

Note that testparm automatically removes base, omitted, and empty levels from the reported constraints, but _ms_extract_varlist shows all levels by default. You can use option noomitted to remove the omitted and empty levels from being posted to r(varlist).

[Caterina Felici]

Thank you very much for your reply! I am trying to extract just two names in from the list of names in local varnames with the extended function ": word of" but I cannot extract multiple words at once. So I tried creating a local macro for the position of the 2 names I need (position 4 and 5 in the list) and used it within the extended function ": word of" as below:

logit .....
local varnames: colfullnames r(table)
local i "4 5"
local ter "`: word `i' of `varnames''"
invalid syntax

cannot find the issue with the syntax though. Or is there a better way of extracting only a few words from the list of the matrix coulmn names?

[Jeff Pitblado (StataCorp)]

The : word # of extended macro function can only get one word/piece at a time.

You can grab the sub matrix then pull the stripe elements from that.
Here is an example.

. sysuse auto
(1978 automobile data)

. regress mpg turn for#rep
note: 1.foreign#1b.rep78 identifies no observations in the sample.
note: 1.foreign#2.rep78 identifies no observations in the sample.

      Source |       SS           df       MS      Number of obs   =        69
-------------+----------------------------------   F(8, 60)        =     11.64
       Model |  1423.05369         8  177.881711   Prob > F        =    0.0000
    Residual |  917.149207        60  15.2858201   R-squared       =    0.6081
-------------+----------------------------------   Adj R-squared   =    0.5558
       Total |   2340.2029        68  34.4147485   Root MSE        =    3.9097

-------------------------------------------------------------------------------
          mpg | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
--------------+----------------------------------------------------------------
         turn |  -.9575443   .1549821    -6.18   0.000    -1.267555    -.647534
              |
foreign#rep78 |
  Domestic#2  |   .3991678   3.112735     0.13   0.898     -5.82723    6.625566
  Domestic#3  |  -1.326173   2.867219    -0.46   0.645    -7.061465     4.40912
  Domestic#4  |  -1.598011   3.060287    -0.52   0.603    -7.719496    4.523473
  Domestic#5  |   6.212278   3.985763     1.56   0.124    -1.760435    14.18499
   Foreign#1  |          0  (empty)
   Foreign#2  |          0  (empty)
   Foreign#3  |  -3.092751   3.675524    -0.84   0.403    -10.44489    4.259391
   Foreign#4  |  -1.856377   3.194688    -0.58   0.563    -8.246705    4.533951
   Foreign#5  |   .1200365   3.170697     0.04   0.970    -6.222301    6.462374
              |
        _cons |   60.25932   6.929618     8.70   0.000     46.39802    74.12062
-------------------------------------------------------------------------------

. matrix list r(table)

r(table)[9,12]
                      0b.foreign#  0b.foreign#  0b.foreign#  0b.foreign#  0b.foreign#  1o.foreign#  1o.foreign#
               turn     1b.rep78      2.rep78      3.rep78      4.rep78      5.rep78     1b.rep78     2o.rep78
     b   -.95754432            0    .39916776   -1.3261725   -1.5980112    6.2122784            0            0
    se    .15498208            .    3.1127355    2.8672192    3.0602866    3.9857629            .            .
     t   -6.1784194            .    .12823697   -.46252917   -.52217699    1.5586171            .            .
pvalue    6.174e-08            .    .89839031    .64537493    .60346973    .12434653            .            .
    ll   -1.2675546            .   -5.8272303   -7.0614648   -7.7194958   -1.7604346            .            .
    ul     -.647534            .    6.6255658    4.4091198    4.5234733    14.184991            .            .
    df           60           60           60           60           60           60           60           60
  crit    2.0002978    2.0002978    2.0002978    2.0002978    2.0002978    2.0002978    2.0002978    2.0002978
 eform            0            0            0            0            0            0            0            0

          1.foreign#   1.foreign#   1.foreign#            
            3.rep78      4.rep78      5.rep78        _cons
     b   -3.0927512    -1.856377    .12003647    60.259317
    se    3.6755237    3.1946885    3.1706968    6.9296176
     t   -.84144503   -.58108233    .03785807     8.695908
pvalue    .40343989    .56336149    .96992652    3.217e-12
    ll   -10.444893   -8.2467055   -6.2223015    46.398018
    ul    4.2593909    4.5339514    6.4623744    74.120616
    df           60           60           60           60
  crit    2.0002978    2.0002978    2.0002978    2.0002978
 eform            0            0            0            0

. matrix sel = r(table)[1...,4..5]

. local ter : colfullnames sel

. di "`ter'"
0b.foreign#3bn.rep78 0b.foreign#4.rep78