st: test difference of two linear combinations that are differences of means

["Radwin, David" <dradwin@rti.org>]

Dear Statalisters,

How do I test whether the difference of two linear combinations is
statistically significant if the linear combinations are differences of
means?

In the contrived example below, I want to test whether the 0.08 absolute
mean difference between the South and West in mean January temperatures
is different than the 8.88 absolute mean difference between the South
and West in July temperatures. 

I can't use -suest- as some earlier posts have suggested because I get
the error message, "impossible to retrieve e(b) and e(V) in A" (error
198).

David
--
David Radwin, Senior Research Associate
Education and Workforce Development
RTI International
2150 Shattuck Ave. Suite 800, Berkeley, CA 94704
Phone: 510-665-8274

www.rti.org/education

. sysuse citytemp, clear
(City Temperature Data)

. mean tempjan, over(region)

Mean estimation                     Number of obs    =     954

           NE: region = NE
    _subpop_2: region = N Cntrl
        South: region = South
         West: region = West

--------------------------------------------------------------
        Over |       Mean   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
tempjan      |
          NE |   27.88537   .2766693      27.34241    28.42832
   _subpop_2 |   21.69437   .3397395      21.02764    22.36109
       South |    46.1456   .6568972      44.85647    47.43473
        West |   46.22539   .7033823      44.84503    47.60575
--------------------------------------------------------------

. lincom [tempjan]West - [tempjan]South

 ( 1)  - [tempjan]South + [tempjan]West = 0

------------------------------------------------------------------------
------
        Mean |      Coef.   Std. Err.      t    P>|t|     [95% Conf.
Interval]
-------------+----------------------------------------------------------
------
         (1) |   .0797906   .9624243     0.08   0.934    -1.808925
1.968506
------------------------------------------------------------------------
------

. mean tempjuly, over(region)

Mean estimation                     Number of obs    =     954

           NE: region = NE
    _subpop_2: region = N Cntrl
        South: region = South
         West: region = West

--------------------------------------------------------------
        Over |       Mean   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
tempjuly     |
          NE |      73.35   .1843789      72.98816    73.71184
   _subpop_2 |   73.46725   .1841403      73.10589    73.82862
       South |    80.9896   .1881789      80.62031    81.35889
        West |   72.10859   .4051957      71.31342    72.90377
--------------------------------------------------------------

. lincom [tempjuly]West - [tempjuly]South

 ( 1)  - [tempjuly]South + [tempjuly]West = 0

------------------------------------------------------------------------
------
        Mean |      Coef.   Std. Err.      t    P>|t|     [95% Conf.
Interval]
-------------+----------------------------------------------------------
------
         (1) |  -8.881006   .4467604   -19.88   0.000    -9.757754
-8.004258
------------------------------------------------------------------------
------

<<winmail.dat>>