Notice: the WebSM website has not been updated since the beginning of 2018.

Web Survey Bibliography

Title A Comparison of Nonresponse Adjustment Methods with the Case Study of HIES
Year 2005
Access date 17.05.2005

The household income and expenditure survey (HIES) conducted by the Korea national statistical office (KNSO) is a monthly diary survey. The response rate of HIES is approximately 70~80%. To compensate 20~30% unit nonresponse of HIES, KNSO is currently using hot-deck imputation method. For imputation, we first categorize both responded and nonresponded household by 16 regions, 4 sorts of working status of household head, and 3 groups of household members (2, 2-4, 5or more), which results in 192(=16*4*3) corresponding cells to impute. Within each cell, we simply reproduce data of a nonresponded household by duplication of all the records of a chosen responded household. A systematic sampling technique is used to select a household for duplication which is based on the sorted household data by the total income. Current imputation method underestimates variance due to duplication of data and leads to the lack of accuracy in estimation. In order to improve the accuracy of estimation, weighting adjustment methods are considered to make up unit nonresponse of HIES. Several methods are recommended in the literature for adjusting weights to compensate for nonresponse. Two weighting adjustment methods, segmentation (or CHAID) and logistic regression methods, are considered in this paper and compared with the current imputation method.

Access/Direct link ISI 2005 homepage
Year of publication2005
Bibliographic typeConferences, workshops, tutorials, presentations