초록 열기/닫기 버튼
This study illuminates some quantitative methodology for missing data in the research of public administration. Based on theories of missing value, this paper explains the importance of statistical treatment of missing data. The paper applies SPSS to a real data set to compare/evaluate the list-wise deletion method, the UM method, the EM algorithm method, and the stochastic regression method. Results clearly indicate that the list-wise deletion method turns out to provide inefficient and even biased estimates. The UM method reveals serious weakness in terms of its curtailed standard errors for estimates. Thus, it can be argued that some other scientific methodologies such as the EM method or regression method should be contrived and utilized. Both of the EM method and regression method could provide a lot more efficient and reliable statistics for some data sets. It should be also noted that they cannot be a remedy for all types of missing data. Researchers should seek the available information on the mechanism of missing data. Then, they can utilize this information in choosing their statistical weapons to treat the problem.
This study illuminates some quantitative methodology for missing data in the research of public administration. Based on theories of missing value, this paper explains the importance of statistical treatment of missing data. The paper applies SPSS to a real data set to compare/evaluate the list-wise deletion method, the UM method, the EM algorithm method, and the stochastic regression method. Results clearly indicate that the list-wise deletion method turns out to provide inefficient and even biased estimates. The UM method reveals serious weakness in terms of its curtailed standard errors for estimates. Thus, it can be argued that some other scientific methodologies such as the EM method or regression method should be contrived and utilized. Both of the EM method and regression method could provide a lot more efficient and reliable statistics for some data sets. It should be also noted that they cannot be a remedy for all types of missing data. Researchers should seek the available information on the mechanism of missing data. Then, they can utilize this information in choosing their statistical weapons to treat the problem.
키워드열기/닫기 버튼
missing value, em algorithm, stochastic regression, list-wise deletion, unconditional mean imputation
