.-
help for ^testres^                                       (STB-40: sg73)
.-

Table making program testres
----------------------------

    ^testres^ model_name test_number, [text describing the null hypothesis]

where "model_name" is the name of the model specified in the immediately
preceding ^modl^ command and "test_number" is a user specified identifier
for the just issued ^test^ or ^testparm^ command.


^testres^ must be used with a preceding ^modl^ command and ^testres^
immediately follows a ^test^ or a ^testparm^ command.


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

    ^testres^ is used in conjunction with the ^modl^, ^test^ (or ^testparm^),
and ^modltbl^ commands to display in output tables the results (p values) of
tests of linear restrictions from estimated models.


Example
-------

In the following example we estimate two different models, and we want to
compare the estimates of the coefficients and t stats for var1 and var2 in
the two models by displaying the estimates in a ^modltbl^ table. Furthermore,
we want to test whether the coefficients on var3 and var4 in each model are
jointly zero, and whether the coefficient of var5 is twice the coefficient of
var6 in the second model.

   . regress depvar var1 var2 var3 var4

   . modl A var1 var2

   . spec  1 "Baseline model"

   . testparm var3 var4

   . ^testres^ A 1 ,var3 = var4 = 0

   . regress depvar var1 var2 var3 var4 var5 var6

   . modl B var1 var2

   . spec  2 "Model 1 plus var5 and var6"

   . testparm var3 var4

   . ^testres^ B 1 ,var3 = var4 =0

   . test var5=2*var6

   . ^testres^ B 2 ,var5 = 2*var6

   . ^modltbl^ ts A B


In this example, each ^modl^ command identified the model and the variables
for which we want the estimates displayed in the ^modltbl^ table, each
^spec^ command includes a description of the particular model above the
^modltbl^ table, and each ^testres^ command identifies the model and the
particular test for which we want results (p values) displayed. In this case
we want the results of a test of the null hypothesis that the coefficients on
var3 and var4 are jointly zero displayed for both model A and model B, and
we are labeling this test as test "1"; hence, the two commands above,

   . ^testres^ A 1        (the results in model A of test #1)

   . ^testres^ B 1        (the results in model B of test #1)

We also specified that we want the results of a test that the coefficient on
var5 is twice the coefficient on var6 displayed (we label this test as
test #2) for model B; hence, the command above,

   . ^testres^ B 2        (the results in model B of test #2)


The end result in the ^modltbl^ table will be headings describing the specifi-
cations of the two models (from the ^spec^ commands), displays of the
coefficient estimates and t stats for var1 and var2 for model A and model B
(from the ^modl^ commands), and displays of the p values on test #1 for
model A and model B, and on test #2 for model B, followed by a description
of the two different null hypotheses which were tested (from the ^testres^
commands).



Author
------

      John H. Tyler                       
      Harvard Graduate School of Education
      email: tylerjo@@hugse1.harvard.edu


Also see
--------

    STB:  STB-40 sg73
 Online:  help for @modl@ (if installed), @modltbl@ (if installed)