st: Re: Binomial Sign Test for Two Dependent Samples

[Mike Lacy <Michael.Lacy@colostate.edu>]

>Date: Fri, 14 Jan 2005 15:18:53 -0700
>From: Thomas Speidel <thomassp@cancerboard.ab.ca>
>Subject: st: Binomial Sign Test for Two Dependent Samples

> Subject: st: Binomial Sign Test for Two Dependent Samples
>
> I am trying to create little program for a "Binomial Sign Test for Two
> Dependant Samples".
> This test is somewhere between a signrank test and a binomial.  The
> computations are fairly simple:
>
> 1. Find the differences among the pairs
> 2. Ignore all null differences
> 3. Count the number of non-zero, non-missing differences (i.e. n)
> 4. Count the number of positive differences (call it x)
> 5. Compute: Binomial(n, k, 0.5)
>

If you would like to use built-in facilities to calculate this test (known 
more commonly, I think, as the McNemar test), note that this it is 
available in -mcc- and -symmetry- command. (See search McNemar)
The documentation and dialog boxes for both of these procedures uses the 
vocabulary of matched case-controls studies, which is quite a parochial way 
to describe this test. That being said, given data in the "wide" 
format,  with x1 and y1 being variables that indicate success/failure on 
two trials for each subject subject, you can obtain the test statistic you 
desire with:

symmetry x1 y1
or
mcc x1 y1

-symmetry- automatically gives the exact (binomial) as well as asymptotic 
p-value


=-=-=-=-=-=-=-=-=-=-=-=-=
Mike Lacy
Fort Collins CO USA
(970) 491-6721 office





*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/