Yang Chi-Chuan, Chen Yi-Hau, Chang Hsing-Yi
Institute of Statistical Science, Academia Sinica, Taipei, 11529, Taiwan.
Center for Health Policy Research and Development, National Health Research Institutes, Miaoli County, 35053, Taiwan.
Stat Med. 2017 Sep 20;36(21):3380-3397. doi: 10.1002/sim.7355. Epub 2017 Jun 2.
Childhood and adolescenthood overweight or obesity, which may be quantified through the body mass index (BMI), is strongly associated with adult obesity and other health problems. Motivated by the child and adolescent behaviors in long-term evolution (CABLE) study, we are interested in individual, family, and school factors associated with marginal quantiles of longitudinal adolescent BMI values. We propose a new method for composite marginal quantile regression analysis for longitudinal outcome data, which performs marginal quantile regressions at multiple quantile levels simultaneously. The proposed method extends the quantile regression coefficient modeling method introduced by Frumento and Bottai (Biometrics 2016; 72:74-84) to longitudinal data accounting suitably for the correlation structure in longitudinal observations. A goodness-of-fit test for the proposed modeling is also developed. Simulation results show that the proposed method can be much more efficient than the analysis without taking correlation into account and the analysis performing separate quantile regressions at different quantile levels. The application to the longitudinal adolescent BMI data from the CABLE study demonstrates the practical utility of our proposal. Copyright © 2017 John Wiley & Sons, Ltd.
儿童期及青少年期超重或肥胖（可通过体重指数（BMI）进行量化）与成人肥胖及其他健康问题密切相关。受儿童与青少年长期发展行为（CABLE）研究的启发，我们对与青少年纵向BMI值的边际分位数相关的个体、家庭和学校因素感兴趣。我们提出了一种用于纵向结果数据的复合边际分位数回归分析的新方法，该方法可同时在多个分位数水平上进行边际分位数回归。所提出的方法将Frumento和Bottai（《生物统计学》，2016年；72:74 - 84）引入的分位数回归系数建模方法扩展到纵向数据，适当地考虑了纵向观测中的相关结构。还开发了针对所提出模型的拟合优度检验。模拟结果表明，所提出的方法比不考虑相关性的分析以及在不同分位数水平上进行单独分位数回归的分析效率更高。将其应用于CABLE研究中的青少年纵向BMI数据证明了我们提议的实际效用。版权所有© 2017约翰威立父子有限公司。
