问卷调查分析中的Logistic回归与自变量筛选问题研究

调查问卷的答卷大多数是既有定性数据又有定量数据,降维和回归都是这种分析的主要手段,而定性数据的降维是难点。本文运用多元分析中的聚类分析和Logistic回归分析,通过聚类的方式既达到了降维的目的,又解决了过离散问题。聚类分析同时还将分散的信息加以集中,使得不同类别的特征更加明显。累积Logistic回归是处理多维多分类定性数据的较好方法,运用逐步回归可以筛选出对研究对象影响最显著的变量,将潜在的信息挖掘出来,为管理决策提供建议和依据。

Research on Logistic Regression and Independent Variables Preparation in Questionnaire

Questionnaires data are both qualitative data and q uantit ative data. Reduction-dimension and Regression are the main analysis method. Red uce dimension on qualitative data is of the difficulty. This paper presents a ne w method to analyze data from questionnaires. The method applies clustering anal ysis and logistic regression with respect of the characteristic of questionnaire s data and special purpose of researches. As an application, we analyzed a surve y data that reflected the staff satisfactions to their own companies, the final result successfully finds out the key factors that influence the staff satisfact ions significantly.