Dropping variables with collinearity

[Philip Hiller]

Hi everyone, I wan to remove all the variables in my data which have collinearity. I found the command "_rmcoll" but it does not automatically drop the collinear variables but only list them.

In the example from the Stata manual:

webuse auto
generate tt = turn + trunk
_rmcoll price-tt, forcedrop

I would like Stata to remove the variable tt.

Apparently by adding "forcedrop" the collinear variables should be dropped, but when I try it, this does not happen. Am I missing something or is there another way to drop that list of variables?

Any help would be very much appreciated!

[Marcos Almeida]

Maybe this text from the Stata Manual helps:

forcedrop specifies that collinear variables be dropped from the variable list instead of being flagged. This option is not allowed when the variable list already contains flagged variables, factor variables, or interactions.

Maybe typing - display r(varlist) - with show what happened.

[Carlo Lazzaro]

Philip:
even with -regress- Stata omits one of the variable creating extreme collinearity issue.
The omitted variable is -turn-:

. webuse auto
(1978 Automobile Data)

.
. generate tt = turn + trunk

. regress price turn trunk tt
note: turn omitted because of collinearity

      Source |       SS           df       MS      Number of obs   =        74
-------------+----------------------------------   F(2, 71)        =      4.91
       Model |  77228695.2         2  38614347.6   Prob > F        =    0.0100
    Residual |   557836701        71  7856854.94   R-squared       =    0.1216
-------------+----------------------------------   Adj R-squared   =    0.0969
       Total |   635065396        73  8699525.97   Root MSE        =      2803

------------------------------------------------------------------------------
       price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        turn |          0  (omitted)
       trunk |   11.75987    169.353     0.07   0.945    -325.9204    349.4401
          tt |   126.6771   93.30743     1.36   0.179    -59.37263    312.7268
       _cons |  -761.7631   3108.734    -0.25   0.807    -6960.404    5436.877
------------------------------------------------------------------------------

[Philip Hiller]

Thank you very much for the quick answers. The thing is that I need the data set without collinear variables for later calculations. Hence the fact that Stata accounts for it when using the reg command does not help me much. However, I think I found what I was looking for which goes in the direction of Marcos' hint. I tried the following command and it worked:

webuse auto
generate tt = turn + trunk
_rmcoll price-tt, forcedrop
keep `r(varlist)'

Best regards,
Philip

[Carlo Lazzaro]

Philip:
the need for removing collinear variables in whatever feasible way may conceal a model misspecification issue.

[Philip Hiller]

Hi Carlo, thank you for the hint. I will take that into account.

Best regards,
Philip