Analyzing "Forest Bats Activities" Data By Principal Component Regression

Regarding the classical models of regression analysis, people generally assume that its response variable is the continual variable. But in the process of the actual problem analysis, we always meet that the response variable possibly descripts variable dispersancy in the way of separated counting. Analyzing this kind of model of the influencing factor of separate counting variable is called the count model(Count Model). To solve this model, there are methods of Poisson regression model, the negative binomial regression model, the Panel method, the maximum likelihood method and so on. And, regression model based on the distributed Poisson is in widespread use in the research of counting model question. Sometimes, the explanatory variable which affect this kind of separate counting variable is so many that inevitably increase the complexity of analyzing the question. How to cause the complex degree of analysis to reduce in the basic guarantee of primitive explanatory variable information foundation ?This article first taking the common dimensionality reduction method of statistical analysis- principal components analysis (PCA) as a foundation, replaces the primitive explanatory variable with the few several aggregate variables, in the guarantee of primitive explanatory variable's information basic invariable foundation, reducing the variable integer, reduces the dimension the high Uygur's explanatory variable. Then combining the dispersancy response variable with these comprehensive explanatory variable, finally establishes model which the response variable forecast. Thus, it namely has guaranteed basic invariable of the primitive explanatory variable information, also causes originally the complex regression analysis process, obtaining certain simplification.