Re: st: How to cluster in 2 x 2 tabulation

[Bo MacInnis <bo@macinnis.org>]

Steve, Thank you much for your insightful comments. You are right that 
clustering by interviewers does not change the point (frequency) 
estimates, but it does change (very slightly) the second-order Rao-Scott 
adjusted F-statistic and p-value, which is of the greatest interest to 
the project.

Interviewer effects? Oh yeah, a common suspect in survey research. It 
would involve some regression, and random effects model sounds good, but 
testing for interviewer effects is not of interest to the particular 
hypothesis in my current paper.  Thank you for your helpful suggestions!! Bo

On 1/18/2011 5:57 PM, Steven Samuels wrote:
> Bo, your question doesn't make sense to me. "Clustering" by 
> interviewer in the -svyset- statement wouldn't change any of the 
> estimates.   It sounds like your Boss wants to know if there are 
> interviewer effects.  I think that some random effects model, perhaps 
> ignoring some of the design, would provide the answer.
>
> Steve
>
> Steven J. Samuels
> sjsamuels@gmail.com
> 18 Cantine's Island
> Saugerties NY 12477
> USA
> Voice: 845-246-0774
> Fax:    206-202-4783
>
> On Jan 18, 2011, at 8:20 PM, Bo MacInnis wrote:
>
> Wow! Thank you loads for your help! The simple solutions like such are 
> particularly appreciated!!!
>
> On 1/18/2011 3:01 PM, Stas Kolenikov wrote:
> > Most likely, all you need is
> >
> > svyset interviewerID
> > svy : tabulate response form
> > svyset interviewerID [pw=weight]
> > svy : tabulate response form
> >
> >
> > On Tue, Jan 18, 2011 at 4:11 PM, Bo MacInnis<bo@macinnis.org>  wrote:
> > > Dear Stata listmates,
> > >
> > > I apologize for asking a possibly extremely simple question, but I have
> > > looked and could not find the answer.
> > >
> > > I have data on a cross-sectional survey with split sample design for 
> > > one
> > > of the questions of interest.
> > >
> > > I need to use the split-sample design (questionnaire form A versus form
> > > B) for a question X, and test whether the response to X 
> > > (dichotomous) is
> > > the same across questionnaire forms using Pearson's test statistics.
> > > However, the boss would like to know the answer when we cluster by
> > > interviewers (interviewer ID) who administered the telephone survey.
> > >
> > > I have two questions:
> > > 1. how would you cluster by interviewer ID in two-way svy:tabulate?
> > > 2. how would you cluster by interviewer ID in two-way tabulate (without
> > > adjusting for sample weight)?
> > >
> > > Thank you for your help!
> > > Bo
> > >
> > > *
> > > *   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/
>
> *
> *   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/
>
> *
> *   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/

*
*   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/