删失数据下部分线性模型的贝叶斯P-样条估计

Bayesian P-splines of partial linear model with censored data

针对右删失数据,建立了部分线性模型.为避免过拟合现象,提出贝叶斯P-样条估计,并将之与B-样条方法进行比较.数值模拟验证了贝叶斯P-样条估计的有效性且该方法降低了节点选择对估计的影响.最终将其运用于卵巢癌生存时间的数据中,发现年龄和治疗方案对卵巢癌影响显著.所得结论对卵巢癌的防治具有重要的实用价值.

For the censored data,established partial linear model.To avoid overfitting,this paper presents the Bayesian P-spline and compares with B-spline method.Simulations studies the validity of Bayesian P-spline is verified and this method reduce the knot selection effect for estimation.Finally,apply them to analyze the factors affecting the survival time of ovarian cancer.In our analysis,age and treatment were found to be significantly affected by data of ovarian cancer.The conclusion has important practice value for prevention and treatment of ovarian cancer.