Testing regression coefficient against a hypothesized value

[Noemi Seng]

Dear Statalist community,
I would like to test whether my regression coefficient is significantly different from the number 2 in order to test one of my hypotheses. Could you tell me what is better here, using the command ttest or test (the latter is rather an F-test, right)?

I would appreciate any help!

[Maarten Buis]

test is the command you want. ttest does not use the regression results to compute its test, it just compares means.

[Noemi Seng]

Thank you very much! Could you also tell me whether this is indeed a t-test then or an F-test just with one linear combination?

[Rich Goldstein]

looking at the results will give you the answer; here is a simple, reproducible example:

. sysuse auto
(1978 automobile data)
r; t=0.01 7:23:52

. regress price i.for c.weight##c.weight turn

      Source |       SS           df       MS      Number of obs   =        74
-------------+----------------------------------   F(4, 69)        =     24.94
       Model |   375388473         4  93847118.3   Prob > F        =    0.0000
    Residual |   259676923        69  3763433.66   R-squared       =    0.5911
-------------+----------------------------------   Adj R-squared   =    0.5674
       Total |   635065396        73  8699525.97   Root MSE        =      1940

------------------------------------------------------------------------------
       price | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
     foreign |
   Domestic  |          0  (base)
    Foreign  |   2731.523    658.636     4.15   0.000      1417.58    4045.466
             |
      weight |  -3.214261   2.354074    -1.37   0.177    -7.910511    1.481989
             |
    c.weight#|
    c.weight |   .0012201   .0003714     3.29   0.002     .0004792    .0019609
             |
        turn |  -265.8238   105.6056    -2.52   0.014    -476.5013   -55.14633
       _cons |   13747.59   4907.183     2.80   0.007     3958.029    23537.15
------------------------------------------------------------------------------
r; t=0.24 7:24:10

. test _b[turn]=-150

 ( 1)  turn = -150

       F(  1,    69) =    1.20
            Prob > F =    0.2766

So, we see it is an F test; however, note that an F-test with 1 df in the numerator is the square of the t-test;