The establishment and application of data cleaning and stratified sampling design for the growth and development data bank among children in elementary school in Taiwan

• Main Authors: Chih-Hung Cheng, 鄭志鴻
• Other Authors: Yiing-Mei Liou
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/36066003704623427639

碩士 === 國立陽明大學 === 臨床暨社區護理研究所 === 98 === Abstract Background: Obesity in children is a growing problem. However, there is a lack of integrated sampling modes in existing census data of elementary school children on health examinations. It results in a growth and development database of elementary school children that has less than 90% response rate and cannot completely represent Taiwan or provide parents, schools and governments with a solid foundation to help promote student health and sports. Purpose: This study explores and analyzes the height, weight and body mass index of Taiwanese elementary school children in the growth and development database of the Ministry of Education and develops automatic error detection procedures and stratified sampling. It also compares the homogeneity of the population characteristics between the sample data and parent data and compares the data of height, weight and body mass index of different ages and genders to understand the applicability of sampling. It finally shows the growth and development data with county and country representativeness in height, weight and body mass index. Study population: Elementary school students in academic year 2008 in public and private schools living in 25 counties in Taiwan. Methods: This is a developmental study of automatic error detection and stratified sampling of databases. It collects the cross-sectional census data of elementary school children on health examination in 2008 academic year from the Ministry of Education. The census data has undergone quality analysis and management. Sampling is conducted after error detection of data. The data of 359 villages and towns in academic year 2008 is divided into seven strata according to the degree of urbanization in cities and counties. This study adopts probability proportional to size (PPS) to sample schools and carry out random sampling in schools. Finally samples with 25 counties as representatives are weighted as country samples. The samples are compared with random sampling followed by county stratification adopted in academic year 2004 - 2006 and samples of all valid data without stratified sampling in the 2007 academic year. The variables include height, weight and body mass index, gender, grade and age. SPSS 17.0 is used for documentation and statistical analysis. Repeated measures and orders are used to detect error. Descriptive analysis, chi-square test, t test, one-way analysis of variance and multivariate analysis are used to indicate the details and differences of variables. Results: The error detection procedure of height, weight and body mass index database of the Ministry of Education is completed. The stratified sampling mode is developed based on the data of Miaoli County, which had more than 90% uploading rate in the 2008 academic year. The data is stratified according to the degree of urbanization in each county. PPS is adopted to sample schools and carry out simple sampling in schools. There is no significant difference between the inspected population characteristics and parent data. The sampling validation is completed. Sampling was carried out on the data of Taipei County, Taichung County, Kaohsiung City and Hualien County in academic year 2008 immediately. The sampling data of each county has passed the homogeneity tests of gender and numbers of grade proportion with parent data. The data is merged into an archive which conforms to the country and the weighted country samples have country representativeness. It indicates the design of stratified sampling mode is successful. The weighted country samples in academic year 2008 are compared with the combined samples using simple random sampling with error detection in 25 counties adopted in academic years 2004, 2005, 2006 and data with error detection but without sampling in academic year 2007. The height and weight of the elementary school students are compared. It indicates and compares the data of height, weight, body mass index and anthropometry between different ages and genders. A growing figure of height, weight and body mass index is found. According to the standard of child obesity issued by the Department of Health, the proportions of underweight, normal, overweight and obese children are 22.6%, 52.8%, 12.6% and 12.1% respectively in elementary school students in academic year 2008. The anthropometry is analyzed using the International Obesity Task Force standard in which the proportions of underweight, normal, overweight and obese children are 2.0%, 69.2%, 19.4% and 9.4%, which is higher than other countries. Key words: sampling, elementary school students, height, weight, body mass index, growth and development databa