Estimation of idealized aging curve of clinical measurements based on semi-longitudinal data

点击翻译按钮获取中文摘要

As for theoretical results, we developed a growth curve model, which is an extension of the model due to Vonesh and Carter (1987, Biometrics), to achieve effective analyzes to unblanced repeated measurements (Ohtaki, 1994, J.J.Biometrics). In the extended growth curve models, we reduced the number of unknown parameters by introducing a family of linear structire into the fixed location and the variance-covariance parameters so that the resultant models can provide higher efficiency and easier interpretation in analysis. It was shown that the family of the extended models contains not only the ordinal growth curve model with non-specified structure for parameters but also many useful growth curve models such as "parallel profile model". We have also developed a computer algorithm and a software on PC for estimating idealized aging curve with repeated observations from semi-longitudinal study. The algorithm consists of four steps : the first step is fitting a regression line for each individual observations to obtain annual changing rate of repeated measurements, the second one is non-parametric smoothing the annual changing rates, the third one is numerical integration of the smoothed annual changing rate for estimating mean struture of aging trend, the last step is analyzing the residuals through growth curve models for estimating the variance-covariance of the random components of repeated measurements. The softare is coded in Fortran 77, and the compiled program is available on PC with MS-DOS ver.5/ver.6. As a pratical application of our method for estimating idealized aging curve, we have studied on the age-related change of clinical measurements such as height, the level of Plasma Choresterol and Urinary Acid in Japanese populations (Kuzuya and Shimokata, 1995, Atherosclerosis ; Ohtaki, Shimokata et al., 1996, Proc.of RIRBM).