毕业论文移动版

摘要：在许多社会研究中，大多数反应变量是分类的而不是连续的。为了弥补线性模型的不足，通常可以采用判别分析、Probit分析、对数线性模型等统计方法，其中最流行的是Logistic回归模型。多年来，Logistic回归模型被用于金融、医药学、环境学等领域并取得显著成果。

本文介绍了Logistic回归分析模型的概况及分类。主要研究模型是二分类因变量Logistic回归模型，对其参数估计、模型评价、回归系数的解释及统计推断进行了详细研究。

具体通过搜集最新的财务指标数据，研究二分类因变量Logistic回归模型在我国上市公司财务困境预警研究中的作用，并利用R软件实现。首先使用K-S检验样本的正态性，从而选择非参数检验中的Wilcoxon秩和检验来检验样本预测变量差异的显著性。然后进行相关性分析以避免因财务指标之间相关性造成的多重共线性问题。对剩余的七个初始预测变量进行逐步回归选择出最佳变量组合。拟合最终模型后评价模型，用AIC进行拟合优度检验，类指标，2015年建模组和2016年检验组分别建立分类表等评价模型预测准确性。回归系数的统计推断，通过Wald检验和参数的置信区间评价模型，并用发生比率解释回归系数。总体上看，本文的预警模型对上市公司财务困境预警较为准确有效。

关键词：Logistic回归；模型评价；财务困境预警；R软件

Abstract：In many social studies, most of the response variables are classified rather than con- tinuous. In order to compensate for the shortcomings of linear models, we usually use dis- criminant analysis, Probit analysis, logarithmic linear model and other statistical methods, one of the most popular models is the Logistic regression model. Over the years, the Lo- gistic regression model has been widely used in financial, medicine science, environmental science and many other fields, which have achieved remarkable results.

This paper introduces the profile and classification of Logistic regression analysis model. The main research model is the two-category dependent variables Logistic regres- sion model, and its parameters estimation, model evaluation, regression coefficient inter- pretation and statistical inference are studied in detail.

In this paper, we study the role of the two-category dependent variables Logistic re- gression model in the research of financial distress early warning in China's listed compa- nies, by collecting the latest financial indicator data and use R. Firstly, we use the K-S test sample's normality, so we select the Wilcoxon rank sum test to test the significance of the differences in the predicted variables, which is an important nonparametric test. And then conduct a correlation analysis to avoid the multi-collinearity problem, which is caused by the correlation between financial indicators. The optimal combination of variables is gained by the gradual regression of the remaining seven initial predictors. After the final model is fitted, we need to evaluate the model. Test the goodness of the model by AIC, and

evaluate the accuracy of model with the Pseudo  and the classification tables, estab-lished by using 2015 model group and 2016 test group. The regression factors are esti-mated by the Wald test and then study the confidence interval of the parameters. Moreover, the regression coefficients are explained by the occurrence rate. On the whole, the early warning model of this paper is accurate and effective in predicting the financial distress of our listed companies.

Keywords: Logistic regression; model evaluation; financial distress warning; R soft- ware

目 录

第一章绪论 1

1.1研究背景及意义 1

1.2国内外研究现状及存在的问题 (责任编辑：qin)