Re: st: -svyset- parameters for survey with multiple waves

[Steve Samuels <sjsamuels@gmail.com>]

No, that design effect is worthless to you.

It only helps you if you do an unweighted analysis on data with no clustering and want to know the standard errors for the weighted analysis.  You can get those standard errors yourself with a weighted analysis.  However, because these SEs ignore the effects of clustering, they are too small.


I suggested that you contact Gallup for better data or information. Do so. You might also look up individual scientists (Gallup and non-Gallup) whose names are on Gallup publications and contact them individually for advice.

If, by the way, you don't know what a design effect is, consult a good sampling text. I happen to like Sharon Lohr, Sampling Design & Analysis, but there are others.


Steve






On Nov 6, 2013, at 5:55 PM, Mohamad Mahmoud wrote:

Many thanks Steve for taking the time to  share your thoughts.
The poll manual lists an average design effects by country by wave as
well as the margin of error (one number only for each country/wave
pair). The manual explains that "the design effect calculation
reflects the weights and does not incorporate the intraclass
correlation coefficients. Design effect calculation: n*(sum of squared
weights)/[(sum of weights)*(sum of weights)]".

I have a weighting variable for each respondent.

Can I use the Design effect and weight to correctly specify the
standard errors and coefficients?

thanks,
Mohamad
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/