[R] Stratified study.

Mark Myatt mark at myatt.demon.co.uk
Tue Nov 6 14:58:48 CET 2001

Maarten van Gelder <mvgelder at vt.edu> writes:

>I am analyzing the results of a survey of the student body on the use of 
>certain technologies. The student body was divided in strata according to 
>two criteria, college and user class. The resulting 24 strata were sampled. 
>Subjects filled out a survey with a number of questions, most of them of 
>the yes-no variety. I created a data.frame with the results for the survey. 
>One row for each subject, one column for each question. Additional columns 
>are for a subject identifier, college, and user class. The 'yes' and 'no' 
>responses were represented by '1' and '0' respectively. The other responses 
>were already numeric.
>The design is that of a stratified random sampling with unequal weights. I 
>created a weight vector 'wt' with the weight for each observation 
>(subject), from the total population, the stratum size, and the number of 
>subjects in each stratum. Using the weighted.mean procedure, I calculated 
>the mean for each question. Per strata means I can calculate with the 
>aggregate function. This is about as far as my knowledge of statistics and 
>of R has taken me. I am not sure how to go about testing for a stratum 
>effect (on a per question basis, or across all questions), or pair wise 
>comparisons between strata (e.g. are there proportionally more laptop 
>owners among the undergraduate engineering students than among their peers 
>in architecture). I would greatly appreciate it if someone could steer me 
>in the right direction.

This looks like a job for GEE methods. These are implemented in the gee
add-in package.


Mark Myatt

r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch

More information about the R-help mailing list