Related Experiment Videos
Improved imputation of non-responses to mailback questionnaires
J W Drane1, D Richter, C Stoskopf
1Epidemiology and Biostatistics Department, University of South Carolina School of Public Health, Columbia 29208.
Statistics in Medicine
|February 1, 1993
Summary
Low survey response rates can be addressed using repeated mailouts and an exponential model. This method allows for the imputation of non-responder data and estimation of population parameters, improving survey validity.
Area of Science:
- Survey methodology
- Statistical modeling
Background:
- Poor survey response rates (10-30%) challenge data validity.
- Understanding non-respondent characteristics is crucial.
Purpose of the Study:
- To develop a statistical model for handling low survey response rates.
- To enable imputation of non-responder data and estimate population parameters.
Main Methods:
- Utilizing repeated mailouts to gather data from resistant individuals.
- Assuming an exponential drop-off in response rates for imputation.
- Applying a second exponential drop-off model for item-specific responses.
Main Results:
- Closed-form estimates for population parameters derived.
- Associated standard errors calculated.
- Model validity tested with three or more mailouts.
Conclusions:
- The proposed exponential model offers a viable method for addressing low response rates in surveys.
- Imputation techniques enhance the validity of survey findings despite non-response.
- Repeated mailouts and modeling improve statistical inference from survey data.