Special feature: theory and practice of surveys
- PDF / 230,965 Bytes
- 2 Pages / 439.37 x 666.142 pts Page_size
- 4 Downloads / 233 Views
Special feature: theory and practice of surveys Hiroshi Saigo1
© Japanese Federation of Statistical Science Associations 2020
It is my great pleasure to introduce to the reader this special feature entitled “Theory and Practice of Surveys”. Survey statistics is one of the most traditional research areas in statistics. Many introductory statistics textbooks explain surveys in detail as a successful example of the application of sampling theory. Survey statistics, however, is still a developing field due to practical challenges: decreasing response rates, outlier detection, and data integration, to name a few. In this special feature, we find seven articles written by leading statisticians about currently studied topics on surveys. In “Empirical likelihood and estimating equations in survey data analysis”, Professors Changbao Wu and Mary E. Thompson review how we combine survey data with statistical models. Traditionally, analysis of survey data is design-based and separated from statistical models. The purely model-based approach, on the other hand, ignores sampling designs in inference. Wu and Thompson provide a unified method of incorporating statistical models in the design-based framework through empirical likelihood and estimating equations. Professors David Haziza and Audrey-Anne Vallée present a comprehensive review on variance estimation for singly imputed survey data. Imputation is a standard technique in handling missing data in surveys. Single imputation, in particular, is most frequently used in official statistics. Variance estimation is challenging due to the variation introduced by imputation. Haziza and Vallée show how we can decompose the variation to make variance estimation feasible, how we can make use of the decomposition in estimating the variance, and how we can compare the methods through a comprehensive simulation study. Professors Shu Yang and Jae-kwang Kim, in their paper entitled “Statistical data integration in survey sampling: A review”, detail a theoretical basis for combining probability and nonprobability samples. Data integration is regarded as the most promising approach to producing statistics in the big-data era. Nevertheless, incorporating nonprobability samples in statistical inference is, of course, technically demanding. Yang and Kim bridge the two kinds of samples through the standard missing data analysis. * Hiroshi Saigo [email protected] 1
Faculty of Political Science and Economics, Waseda University, 1‑6‑1 Nishiwaseda, Shinjuku, Tokyo 169‑8050, Japan
13
Vol.:(0123456789)
Japanese Journal of Statistics and Data Science
The three articles above were presented at an international session in the 2020 Joint Statistical Meeting of the Japanese Federation of Statistical Science Association entitled “Theory and Practice of Surveys”, sponsored by the Japan Society of the Promotion of Science KAKENHI Grant Number 19HP2005 for enhancing international recognition of JJSD. Professors Masaki Mitsuhiro and Takahiro Hoshino propose a new data integration method which they cal
Data Loading...
