Equivalence tests of means                             [(STB-17: sg21, sg21.1]
--------------------------

        ^equim^ varname ^=^ # [^if^ exp] [^in^ exp] [, ^d^elta(str) ^sd^elta(str)
                                           ^r^andom]

        ^equim^ varname ^=^ varname [^if^ exp] [^in^ exp] [, ^d^elta(str) ^sd^elta(str)
                                           ^r^andom]

        ^equim^ varname [^if^ exp] [^in^ exp] [, ^d^elta(str) ^sd^elta(str)
                                           ^r^andom ^b^y(varname)]

        ^equimi^ n1 mean1 sd1 benchmean [, ^d^elta(str) ^sd^elta(str) ^r^andom]

        ^equimi^ n1 mean1 sd1 n2 mean2 sd2 [, ^d^elta(str) ^sd^elta(str) ^r^andom]


tests the null hypothesis that the means differ by at least ^delta^ (or, by
^delta^ in the negative direction and ^sdelta^ in the positive direction).  
In the third form, ^equim^ tests that the mean of varname is equivalent 
within the two groups specified in ^by^(varname).

^equimi^ is the immediate version of ^equim^.  In the second form of the 
command, n1, mean1 and sd1 should be the values for the control, baseline, 
or reference group.

Options
-------
^delta^ and ^sdelta^ control the width of the equivalence interval.  If 
    neither option is specified, a default of .2*mean1 is used for both
    ^delta^ and ^sdelta^.  In the first form of the command, however, ^delta^ 
    is 20 percent of this number.  ^delta^ can be entered either as a real 
    number or as a percent.  If you enter it as a percent, then you ^must^ 
    include the "%" symbol.

^random^ indicates that ^delta^ is to be treated as an estimated amount.  As 
    a consequence, the variance is increased (i.e., the confidence intervals
    are widened).  If ^random^ is not specified, ^delta^ is treated as fixed 
    and known.


Author
------

        Richard Goldstein
        EMAIL  richgold@@netcom.com
        


Also see
--------

    STB:  sg21 (STB-17), sg21.1 (STB-24)
On-line:  ^help^ for @equip@ if installed