re: st: Should rim weights be taken as sampling weights in SVYSET, say?

[Steven Samuels <ssamuels@albany.edu>]

A cursory web search shows that 'rim weighting" is a synonym for  
sample balancing by iterative proportional fitting, and is also known  
as "raking", or "ratio raking".  I 've never seen the term in a text  
book, but it
appears to be in common use. Raking is a procedure for adjusting the  
sample weights so that the sample marginal proportions match the  
target population marginal proportions on a number of
categorical characteristics at the same time:   age, gender,  
race..... .  These weights are constructed by starting with the  
sampling weights and doing further adjustments.

This procedure removes much non-response or sampling bias arising  
from failure of the sample to reflect these population  
characteristics.  It does not NOT remove biases
due to differences between responders and non-responders unrelated to  
age, gender... race.


Use of these weights is standard in survey sampling--so go ahead and  
use them.


Note: Post-stratification in STATA as implemented adjusts for only  
one characteristic and, so, is not an implementation of a raking  
adjustment. However Nick Winter at Cornell has written an excellent  
raking program   'SURVWGT,'  which can be found by a "search, all."

steve


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/