首页> 外文学位 >A COMPARATIVE STUDY OF METHODS OF ESTIMATING THE PARAMETER IN A LINEAR LOGISTIC REGRESSION MODEL FOR BINOMIAL RESPONSE DATA.
【24h】

A COMPARATIVE STUDY OF METHODS OF ESTIMATING THE PARAMETER IN A LINEAR LOGISTIC REGRESSION MODEL FOR BINOMIAL RESPONSE DATA.

机译:二项式响应数据的线性物流回归模型中参数估计方法的比较研究。

获取原文
获取原文并翻译 | 示例

摘要

To estimate the parameter in a linear logistic regression model, various alternative methods have been suggested. A comparison is made among maximum likelihood, minimum Pearson's chi square, minimum Neyman's chi square, minimum modified logit chi square (which includes Berkson's minimum logit chi square as a special case), bias corrected maximum likelihood, bias corrected Berkson's minimum logit chi square, and Rao-Blackwellized Berkson's minimum logit chi square in terms of three properties of the associated estimators. These properties are (1) the behavior of the estimator under the usual fixed dimension asymptotics where the number of values of the independent variable vector under consideration is fixed while the number of observations at each value of the independent variable vector gets large, (2) the first order approximations to the bias, the variance, and the mean squared error of each component of the estimator, and (3) the behavior of the estimator under increasing dimension asymptotics where the number of values of the independent variable vector under consideration gets large as the total number of observations gets large in such a way that their ratio remains bounded away from zero. On the basis of this comparison, some general recommendations are given concerning which methods to use and which methods not to use in a given situation. For example, when the number of values of the independent variable vector under consideration is "large" while the average number of observations at each value of the independent variable vector is "small", we recommend using either the method of maximum likelihood or the method of bias corrected maximum likelihood since these two methods produce estimators which are consistent under increasing dimension asymptotics while the other methods under consideration do not.
机译:为了估计线性逻辑回归模型中的参数,已经提出了各种替代方法。在最大似然,最小皮尔逊卡方,最小内曼卡方,最小修改对数卡方(特殊情况下包括伯克森最小对数卡方),偏倚校正的最大似然,偏向校正伯克森的最小对数卡方之间进行比较,和Rao-Blackwellized Berkson的最小对数卡方,这是根据相关估计量的三个属性得出的。这些特性是(1)估计器在通常的固定维渐近性下的行为,其中所考虑的独立变量矢量的值数量是固定的,而独立变量矢量的每个值的观测值数量变大,(2)估计量各分量的偏差,方差和均方误差的一阶近似值,以及(3)渐近变数渐近情况下估计量的行为,其中所考虑的独立变量向量的值数量变大随着观察总数的增加,其观察数的比率仍保持在零附近。在此比较的基础上,给出了有关在给定情况下使用哪些方法和不使用哪些方法的一些一般性建议。例如,当所考虑的自变量向量的值的数量为“大”而自变量向量的每个值的平均观察次数为“小”时,我们建议使用最大似然法或偏差校正的最大似然估计是因为这两种方法产生的估计量在维数渐近渐近的情况下是一致的,而正在考虑的其他方法则不然。

著录项

  • 作者

    DAVIS, LINDA JUNE.;

  • 作者单位

    Rutgers The State University of New Jersey - New Brunswick.;

  • 授予单位 Rutgers The State University of New Jersey - New Brunswick.;
  • 学科 Statistics.
  • 学位 Ph.D.
  • 年度 1983
  • 页码 327 p.
  • 总页数 327
  • 原文格式 PDF
  • 正文语种 eng
  • 中图分类
  • 关键词

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号