Factor variables with collapse

[Paul Cauley]

I am having difficulty with factor variables (i.variable) with a collapsed dataset. The collapse command worked without a problem and I can do other analysis without issues.


But when I try to do a regression using the i. prefix, it gives the error "factor variables not allowed". If I used the interaction notation ##, I get "interaction variables not allowed"

For example: reg income i.city


Is it not possible to use factor notation with a collapsed dataset?

[Nick Cox]

I think that depends on the collapse command and the dataset. Can you give a reproducible example?

[Paul Cauley]

This gives the same error as I am getting: factor variables not allowed

sysuse auto

collapse mpg, by( rep78)

reg  mpg i.rep78

[CODE]

Without the i. prefix, it doesn't give the error.

reg  mpg rep78

[Nick Cox]

This works (namely, is legal syntax) for me in Stata 15.1. I suppress detail peculiar to my computer and licence.

. sysuse auto, clear
(1978 Automobile Data)

. collapse mpg, by(rep78)

. regress mpg i.rep78

      Source |       SS           df       MS      Number of obs   =         5
-------------+----------------------------------   F(4, 0)         =         .
       Model |  44.3347197         4  11.0836799   Prob > F        =         .
    Residual |           0         0           .   R-squared       =    1.0000
-------------+----------------------------------   Adj R-squared   =         .
       Total |  44.3347197         4  11.0836799   Root MSE        =         0

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       rep78 |
          2  |     -1.875          .        .       .            .           .
          3  |  -1.566668          .        .       .            .           .
          4  |    .666666          .        .       .            .           .
          5  |   6.363636          .        .       .            .           .
             |
       _cons |         21          .        .       .            .           .
------------------------------------------------------------------------------

. about

Stata/SE 15.1 for Windows (64-bit x86-64)
Revision 08 May 2018
Copyright 1985-2017 StataCorp LLC

In general, a regression like this would not be helpful, as it merely expresses a tautology, and reduces to saying that the mean for each category is the mean for each category.

But the code is legal and I can't reproduce the error message that you're claiming.

If you're using a version earlier than 15.1 you're asked to tell us that.

[Paul Cauley]

Ah my apologies - I was using Stata 12. I just tried it on another computer with Stata 13, and it works.

Thanks!

[Nick Cox]

Thanks for the closure. Factor variable notation was introduced in Stata 11, but I recollect, or imagine, that it wasn't rolled out to all relevant commands until later.