Discussion of sampling bias in estimating concordance correlation coefficients for longitudinal normal data

• Main Authors: Lee,Chia-Yin, 李佳音
• Other Authors: Tsai,Miao-Yu
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/tfc9yb

碩士 === 國立彰化師範大學 === 統計資訊研究所 === 107 === In clinical studies, concordance correlation coefficient (CCC) is the most popular measures of agreement for a continuous scale. However, sample selection bias has long been recognized in many fields including clinical trials, epidemiology studies, genome-wide association studies, and wildlife management area. Therefore, the focus of this thesis is on evaluating how subject distribution and sample size affect the estimation of CCC under the existence of sampling bias. We use the variance components (VC) method to assess the intra-, inter- and total-CCCs and compare the performance of VC and U-Statistics (US) methods for linear mixed models (LMM). Simulation results show that CCC from the non-normal distribution is overestimated. Hence, we proposed a resampling procedure to reduces the influence of subject distribution on the estimation of CCC. Finally, an application of corticospinal diffusion tensor tractography study is used for illustration.