Statalist The Stata Listserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: non-response weights

From   Andrew Mitchell <>
Subject   st: non-response weights
Date   Fri, 5 May 2006 11:37:25 -0400

I am attempting to create weights to account for non-response and am wondering if my approach seems correct. Non-response to my outcome variable of interest occurred because potential respondents were not present in a given household when interviewers came by (all household members present were interviewed). However, I do have socio-demographic data on those respondents - such as years of education, age, marital status, gender - which, in turn, appear to be related to whether or not respondents were present (a quick-and-dirty logistic regression with "not present" as the outcome indicated significant socio-demographic relationships). I believe that this situation lends itself to non-response weighting in my regressions, but I have a couple of questions I'm hoping someone can clarify:
1) the svy/survwgt package appears to create non-response weights by re-calibrating sample design weights. My dataset does not include sample design weights because every person present in a given household was interviewed (or proxied for in the case of children). Assuming that the sample represents a random selection of households, is there any problem with assigning everyone a sample design weight of "1"? The non-response re-calibration would then adjust in relation to that (constant) weight.
2) can the resulting weight generated by the "survwgt nonresponse" command entered directly into the svyset command (e.g., "svyset [pweight=nonresponsewgt], strata(householdid)")?

Thanks for any guidance people can provide.

Andrew Mitchell

* For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index