Online Program

Return to main conference page

All Times ET

Program is Subject to Change

Tuesday, June 15
Tue, Jun 15, 11:30 AM - 1:00 PM
TBD
Treatments for Nonresponse: Imputation, Weighting, and Other Models

Balanced Imputation for Swiss Cheese Nonresponse (308160)

*Esther Eustache, University of Neuchatel 
Yves Tillé, University of Neuchatel 
Audrey Anne Vallee, University Laval 

Keywords: calibration; non-monotone nonresponse; random imputation; variance

Swiss cheese nonresponse, also known as non-monotone nonresponse, occurs when all variables of a survey contain missing values without a particular pattern. The estimators of the parameters of interest can be significantly affected by the missing values, which leads to a bias and an increase variability. To reduce the effects of nonresponse, the missing values are usually imputed. When several variables of a dataset need to be imputed, it may be difficult to preserve the distributions and the relations between the variables. A new method using random imputations by donors will be presented. It extends the balanced K-nearest neighbor imputation to the treatment of Swiss cheese nonresponse. The construction of the method meets the following requirements. First, a nonrespondent should be imputed by neighboring donors. Next, the same donor should impute all the missing values of the same nonrespondent. Finally, donors are selected to meet the balancing constraints to eliminate the variance due to imputation. To meet all the requirements, a matrix of imputation probabilities is constructed using calibration techniques. The donors are then selected with these imputation probabilities and balanced sampling methods. The method will be implemented in the "SwissCheese’’ R package.