Jen Zhen <[email protected]> :
You've gotten advice on testing equality of means/proportions one by
one, but let me urge you to do a joint test. Categorical variables
can be broken into dummy variables (all but one category turned into
that many dummies), and then you can use -hotelling- to test equality
of all relevant means (AKA proportions for dummy variables) across
treatment and control. This is commonly called a test of balance, and
Hotelling's test can also be done using a regression of a treatment
dummy on all variables whose means should be equal across groups,
which suggests natural generalizations using e.g. various robust VCEs:
http://www.stata.com/statalist/archive/2009-02/msg01191.html
On Wed, Jan 27, 2010 at 10:34 AM, Jen Zhen <[email protected]> wrote:
> Hi there,
>
> I have a summary table with members from (1) the treatment group and
> (2) the control group of an experiment, and would like to provide for
> each observable variable a formal test for whether the two sets of
> people come indeed from the same population. For the continuous normal
> variables, I am just doing a t-test to test for equality of means, but
> I'm not fully sure what to do instead for the categorical variables
> (with more than 2 categories).
>
> I found the -csgof- command for a Chi-Square test of whether the
> proportions of the different categories in one sample correspond to
> those coming from some hypothesis, but which command would you use to
> test whether the proportions in the two groups are identical?
>
> Thanks,
> J
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
