.-
help for ^svytest^
.-

Test multidimensional linear hypotheses after ^svy^ estimation commands
---------------------------------------------------------------------

    (1)  ^svytest^ exp ^=^ exp [^,^ ^noadj^ust ^a^ccumulate ^no^test ]

    (2)  ^svytest^ coefficient_list [^,^ ^noadj^ust ^a^ccumulate ^no^test ]

    (3)  ^svytest^ [varlist] [^,^ ^bon^ferroni ]


exp ^=^ exp is a linear equation that is valid for the ^test^ command;
coefficient_list is a valid coefficient_list for the ^test^ commands;
see help @test@ and [5s] test.


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

^svytest^ tests multidimensional linear hypotheses after a ^svy^ estimation
command; namely, ^svymean^, ^svytotal^, ^svyratio^, ^svyreg^, ^svylogit^, or ^svyprobt^.

Syntax (1) allows you to build up a multidimensional hypothesis consisting
of any number of linear equations.  If you simply want to estimate a single
linear combination, use the ^svylc^ command; see help @svylc@.

Syntax (2) gives a test of hypotheses of the form x1=0, x2=0, x3=0, etc.

Syntax (3) is only available after the regressions ^svyreg^, ^svylogit^, and
^svyprobt^.  It computes a Bonferroni adjustment for hypotheses of the form
x1=0, x2=0, x3=0, etc.

By default, ^svytest^ used with syntax (1) or (2) computes an adjusted Wald
test: (d - k + 1)*W/(k*d) ~ F(k, d - k + 1), where W is the Wald test
statistic, k is the dimension of the hypothesis test, d = total number of
sampled PSUs minus the total number of strata, and F(k, d - k + 1) is an F
distribution with k numerator d.f. and d - k + 1 denominator d.f.


Options
-------

^noadjust^ specifies that the model Wald test be computed as W/k ~ F(k, d)
    (notation described above).  This gives the same result as the ^test^ 
    command does.

^accumulate^ allows a hypothesis to be tested jointly with the previously tested
    hypotheses.

^notest^ suppresses the output.  This option is useful when you are interested 
    only in the joint test of a number of hypotheses.

^bonferroni^ can be specified only after estimating a model with ^svyreg^,
    ^svylogit^, or ^svyprobt^.  When this option is specified, ^svytest^ displays
    adjusted p-values for each of the coefficients in the varlist.  Adjusted
    p-values are computed as p = min(1, k*invt(d, t)), where k is the number of
    terms in the varlist, t = _b[x]/_se[x] is the standard t statistic, and
    d = total number of sampled PSUs minus the total number of strata.  If
    no varlist is specified with the ^bonferroni^ option, adjustments are 
    made for all terms in the model excluding the constant.


Examples
--------

 . ^svyreg y x1 x2 x3 x4^
 . ^svytest x3 x4^
 . ^svytest x3 x4, noadjust^
 . ^svytest x3 x4, bonferroni^

 . ^svymean result1 result2 result3^
 . ^svytest result1 result2 result3^
 . ^svytest result1 result2 result3, noadjust^

 . ^svylogit disease gender age1 age2 age3 age4^
 . ^svytest age1 = age2^
 . ^svytest age2 = age3, accum^
 . ^svytest age3 = age4, accum^

If the variables specified with the ^svymean^ , ^svytotal^, or ^svyratio^ command
contain missing values, you will have to use the ^complete^ option when you
first run the command:

 . ^svymean result1 result2 result3, complete^
 . ^svytest result1 = result2, notest^
 . ^svytest result2 = result3, accum^

(Note that ^svyreg^, ^svylogit^, and ^svyprobt^ always use only complete cases.) 

When you have ^by()^ subpopulations, ^svytest^ uses the syntax used by the
^test^ command with multiple equations:

 . ^svymean vitaminc, by(sex)^
 . ^svytest [vitaminc]Male = [vitaminc]Female^

Results for the above hypothesis can also be obtained using the ^svylc^ command:

 . ^svylc [vitaminc]Male - [vitaminc]Female^

The ^svylc^ produces much more output than ^svytest^ and should be used for
computing single linear combinations.  For multiple linear combinations,
^svytest^ should be used:

 . ^svymean vitaminc, by(race)^
 . ^svytest [vitaminc]White = [vitaminc]Black^
 . ^svytest [vitaminc]White = [vitaminc]Other, accum^


Also see
--------

 Manual:  [5s] test
On-line:  help for @svylc@, @svymean@, @svyreg@