Related Experiment Videos
On weighting the rates in non-response weights
Roderick J Little1, Sonya Vartivarian
1Department of Biostatistics, University of Michigan, 1420 Washington Heights, Ann Arbor, MI 48109-2029, USA. rlittle@umich.edu
Statistics in Medicine
|April 22, 2003
Summary
Weighting survey response rates using sampling weights is often incorrect or unnecessary. The correct method models non-response using design variables for accurate survey estimation.
Area of Science:
- Statistics
- Survey Methodology
Background:
- Sample surveys commonly use weights inversely proportional to selection and response probabilities.
- Response weights are typically estimated using the inverse of sample-weighted response rates within adjustment cells.
Purpose of the Study:
- To evaluate the validity of using sample-weighted response rates for adjusting survey estimates.
- To propose a more accurate method for estimating response weights in sample surveys.
Main Methods:
- Simulations were conducted to assess the impact of weighting response rates by sampling weights.
- A modeling approach was proposed, treating non-response as a function of adjustment cells and design variables.
Main Results:
- Weighting response rates by sampling weights leads to biased estimates when design variables correlate with non-response.
- This weighting method is unnecessary when design variables are unrelated to non-response.
Conclusions:
- The proposed method of modeling non-response provides more accurate survey estimates.
- Response propensity weighting is a viable alternative when extensive cross-classification is infeasible.