The heteroscedasticity impact in constructing age-specific reference range

• Main Authors: Hsin-Yi Huang, 黃欣儀
• Other Authors: Chong-Yau Fu
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/52655876953171811296

碩士 === 國立陽明大學 === 公共衛生研究所 === 95 === The reference range of medical data is commonly used in reading biomedical measurement. When the measurement varies with age, an age-specific reference range is required. The construction of reference range is based on statistical concerns, which may be non-parametric or parametric approach. In parametric approach, the regression model is an easy access method, which the age effect may be removed and base on the residual part for further constructing reference range. But, when the data variation increases with age, it will cause a heteroscedasticity problem in statistics. In this study, the impact of heteroscedasticity in constructing age-specific reference range will be investigated. The four adjusting methods, weighted least squares, and the methods referred by Goldstein, H. (1972), Royston, P. (1991) and Altman, D.G. (1993), will be compared. Two data sets, fetal data and prostate-specific antigen (PSA), are used for illustrations.