研究生医学统中计logistic回归.ppt

Logistic regression Logistic回归 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 第一节.非条件logistic回归 第二节.条件logistic回归 第三节. 应用及其注意事项 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 医学研究中常碰到应变量的可能取值仅有两个（即二分类变量），如发病与未发病、阳性与阴性、死亡与生存、治愈与未治愈、暴露与未暴露等，显然这类资料不满足多元（重）回归的条件 什么情况下采用Logistic回归 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. Brown(1980)在术前检查了53例前列腺癌患者，拟用年龄(AGE)、酸性磷酸酯酶(ACID)两个连续型的变量，X射线(X_RAY)、术前探针活检病理分级(GRADE)、直肠指检肿瘤的大小与位置(STAGE)三个分类变量与手术探查结果变量NODES（1、0分别表示癌症淋巴结转移与未转移 ）建立淋巴结转移的预报模型。 实例 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. （一）53例接受手术的前列腺癌患者情况 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. （二）26例冠心病病人和28例对照进行病例?对照研究 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 26例冠心病病人和28例对照者进行病例?对照研究 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 一、logistic回归模型 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 概率预报模型 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 二、模型的参数估计 Logistic回归参数的估计通常采用最大似然法(maximum likelihood，ML)。最大似然法的基本思想是先建立似然函数与对数似然函数，再通过使对数似然函数最大求解相应的参数值，所得到的估计值称为参数的最大似然估计值。 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 参数估计的公式 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile 5.2.0.0. Copyright 2004-2011 Aspose Pty Ltd. 三、回归参数的假设检验 E