Cook-Weisberg test for heteroskedasticity                       (STB-10: srd14)
-----------------------------------------

                       ^cwhetero^ varlist

Description
-----------

^cwhetero^ is a test for heteroskedasticity (nonconstant variance) in a regres-
sion that is a function of one or more of the variables in the regression.  
The varlist is a subset of the right-hand-side variables from the original  
regression (the "subset" may include all the original variables).

This command can only be used after ^fit^.

No options other than the variables you think may be related to a possible
nonconstant variance are allowed.  Your data set will automatically be reset
to the data set on disk.  Thus, if you want to test more than one set of
variables you must re-estimate your original regression (see below).  Remember
that if you want to re-estimate your model without seeing the output, use
"quietly fit ...".  In this way, you can perform several heteroskedasticity
tests and have the results all on the screen at once.

A test, with p-value, of the hypothesis is presented (if test is statistically
significant, there is a problem) and a graph is also presented (if the graph
shows a "wedge" shape there is a problem).  A second test is always presented:
whether the form of the heteroskesticity is a function of the predicted
values, one of the common expectations.  The following example uses the
weisgas.dta data file (S.  Weisberg, Applied Linear Regression, 2d ed.,
NY:  Wiley, 137-140 (discussion starts on p. 135).  The graph is NOT saved.


Example
-------

. us weisgas

. fit y x1-x4

  Source |       SS       df       MS                  Number of obs =      32
---------+------------------------------               F(  4,    27) =   84.54
   Model |  2520.27241     4  630.068101               Prob > F      =  0.0000
Residual |  201.227595    27  7.45287388               R-square      =  0.9261
---------+------------------------------               Adj R-square  =  0.9151
   Total |     2721.50    31  87.7903226               Root MSE      =    2.73

------------------------------------------------------------------------------
       y |      Coef.   Std. Err.       t     P>|t|       [95% Conf. Interval]
---------+--------------------------------------------------------------------
      x1 |  -.0286089   .0906015     -0.316   0.755      -.2145079    .1572901
      x2 |   .2158169    .067718      3.187   0.004        .076871    .3547628
      x3 |  -4.320052   2.850967     -1.515   0.141      -10.16975     1.52965
      x4 |    8.97489   2.772632      3.237   0.003        3.28592    14.66386
   _cons |   1.015018   1.861308      0.545   0.590      -2.804072    4.834107
------------------------------------------------------------------------------

. cwhetero x1

Score test (x1) =    1.395; chi-square p-value (df) =  0.238 (1)
Score test (pred fit) =    0.000; chi-square p-value (df) =  0.985 (1)

. qui fit y x1-x4

. cwhetero x4

Score test (x4) =    0.010; chi-square p-value (df) =  0.922 (1)
Score test (pred fit) =    0.000; chi-square p-value (df) =  0.985 (1)

. qui fit y x1-x4

. cwhetero x1 x4

Score test (x1 x4) =    9.283; chi-square p-value (df) =  0.010 (2)
Score test (pred fit) =    0.000; chi-square p-value (df) =  0.985 (1)

. qui fit y x1-x4

. cwhetero x1-x4

Score test (x1 x2 x3 x4) =   10.299; chi-square p-value (df) =  0.036 (4)
Score test (pred fit) =    0.000; chi-square p-value (df) =  0.985 (1)

Author
------

Richard Goldstein, Qualitas, Brighton, MA  EMAIL goldst@@harvarda.bitnet