Why don't you simply partition the chi-square and test for specific subtables?
Granted, there isn't a button you can push to generate all possible subtables,
but I'm sure you don't want to do that anyway.
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]]On Behalf Of Hoang Thanh
> Huong
> Sent: Tuesday, May 25, 2004 7:11 PM
> To: [email protected]
> Subject: Re: st: t-test command for simple cross-tabs
>
>
> Dear all,
>
> When comparing two proportions from the same survey, I first use command
> "svyset" that taking survey design into account. After that I execute a
> command "svytab X Y, per se".
>
> Hope this works.
>
> Best,
> Hoang Thanh Huong
>
> ----- Original Message -----
> From: "Nick Cox" <[email protected]>
> To: <[email protected]>
> Sent: Wednesday, May 26, 2004 3:23 AM
> Subject: RE: st: t-test command for simple cross-tabs
>
>
> > There were two interesting articles in TAS which may be relevant.
> >
> > Comparing Two Proportions from the Same Survey.
> > C.J. Wild; G.A.F. Seber
> > The American Statistician 47 (Aug. 1993), 178-181.
> >
> > Difference of Proportions from the Same Survey.
> > A.J. Scott; G.A.F. Seber
> > The American Statistician 37 (Nov. 1983), 319-320.
> >
> > As I recall, however, neither boils down to a t-test.
> >
> > I was very interested in these papers when I read them,
> > but never focused on how to do either in Stata. Irrespective
> > of whether either is the answer to Claudia's question,
> > I would be happy to learn of Stata implementations.
> >
> > Nick
> > [email protected]
> >
> > Claudia Solari
> >
> > > Thank you so much for your response. It is my understanding that a
> > > chi-square test can tell me if the relationship between two
> > > variable is
> > > significant, lets say region (ne, mw, w, s) and age cateories (19-29,
> > > 30-34, 35+). The chi-square refers to whole variables, and
> > > the t-test is
> > > based on the categories, such as proportion in a cell (means or
> > > percentages). So, I might want to know if there are significant
> > > differnences in age for those who live in the NE versus the South.
> > >
> > > I know of a free software package that calculates my t-tests
> > > (STATS), but
> > > I'm not sure how to make STATA do it. I was unable to reach the url
> > > that you indicated, so I am not sure if my email is redundant. If you
> > > have any further comments, I would love to hear them.
> >
> > Richard Williams
> >
> > > > I'm probably showing my ignorance here, but what do you want to do a
> > > > t-test
> > > > of, and how would it differ from what the chi-square test
> > > tells you? In
> > > > the special case of 2 X 2 tables, it is possible to do a
> > > z-test using the
> > > > -prtest- command; see p. 7 of
> > > > http://www.nd.edu/~rwilliam/stats1/Categorical-Stata.pdf. Is that
> > > > what you
> > > > want, or do you have something else in mind? If
> > > occupational status is a
> > > > continuous variable, you could of course just use the
> > > t-test command.
> >
> > Claudia Solari
> >
> > > > >I am just trying to find out the command to conduct a
> > > t-test for a simple
> > > > >crosstab, such as gender by occupational status. I know
> > > there is a simple
> > > > >command for the chi-square test, wihc is just an option
> > > for the tabulate
> > > > >command, but I have not found an option for a t-test.
> >
> > *
> > * 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/
> >
>
> *
> * 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/
>
>
*
* 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/
