A Study on Double Frame and Raking with Cross-tables Structure for Longitudinal Data

• Main Authors: LIEN, CHI-HSIUNG, 連啟雄
• Other Authors: WANG, HONG-LONG
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/18894777922606732900

碩士 === 國立臺北大學 === 統計學系 === 104 === Raking is often applied to adjustify sample structure approaching to population structure. Traditional raking method can only fit the sample structure marginaliy (eg: gender, age…etc) and usually the cross-structure can not be preserved. In this study, we were raking sample with cross-structure to make sure the unanimous structure between sample and population. In this study we used the longitudinal data (RI1999, RI2000 and RII2000) collected by PSFD in 1999 and 2000. For summarizing the datasets from two independent years, we took the double-frame cross-structural adjustment approach proposed by Skinner in 1991, which modified from Bankier’s multiple frame method (1986). Along with three different raking methods for cross-structured data, we proposed two pipelines and made some discussion on the weight-effect in order to figure out which is the optimal method. The results indicated that the raking operation apparently altered the analyses but there’s no significant difference between the two methods proposed by us. Based on the our studied, two methods performed similary other on the effectiveness and the speed of convergence, however, the method 2 is slightly better in the figurest.