Re: st: chi2 test for trend with survey data

[Steven Samuels <sjhsamuels@earthlink.net>]

The chi square test for trend in proportions, due to Armitage, is  
formally equivalent to a linear regression of the response on the  
group  scores, with a SE for beta based on the null hypothesis. For  
the closest equivalent in Stata, use -svy: reg- and report the F test  
result.


By the way: In Stata, the chi square test for trend is implemented in  
-tabodds-. -nptrend- gives similar results, but accommodates the 0-1  
response by adding corrections for tied ranks; it is really intended  
for continuous data.


-Steve
On Feb 5, 2009, at 2:25 AM, rafaelcla wrote:

> Dear Listers,
>
> I want to test the proportion of smokers in four different groups  
> for linear trend (chi2). Usually I solve this using "nptrend" but  
> this time I'm dealing with survey data (and so, sampling weights).  
> How to proceed with this test in this data??? (should I just use  
> logistic regression???)
*
*   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/