论文部分内容阅读
为了对医学纵向数据进行建模研究,利用变系数模型探索时变的检测指标与时变的肿瘤大小之间的映射关系,提出基于B样条估计与adaptive-LASSO惩罚最小二乘结合的两步迭代法进行系数估计与变量选择,该方法在系数估计的同时分离变系数、常系数和零系数。将变量选择方法应用于医学肿瘤数据中,对各个变量进行系数估计,研究不同时刻的症状对肿瘤大小进展的影响,并最终对肿瘤大小进行预测。平均相对误差的计算结果表明该模型模拟效果较好。
In order to study the modeling of medical longitudinal data and explore the relationship between the time-varying detection index and the time-varying tumor size by using the variable coefficient model, a two-step approach based on B-spline estimation and adaptive-LASSO penalty least squares Iterative method for coefficient estimation and variable selection, the method of variable coefficient at the same time separate the variable coefficient, constant coefficient and zero coefficient. The variable selection method is applied to the medical tumor data, and the coefficients of each variable are estimated to study the influence of the symptoms at different times on the progress of the tumor size, and finally the tumor size is predicted. The calculation results of average relative error show that the model is better.