The Study And Application Of Model Selection And Model Average Methods

Model selection and model average methods are preceding issues studied by currentstatistics,these two methods have very broad application prospects,and therefore have a very important research significance.Model selection and model average methods are proposed early time,and have been attracting much attention of scholars,for various reasons,related research on the two methods have been comparatively slow.By comparison,relevant theoretical study in model selection method is much more than model average method.Model selection and model average methods cover a broad range of research interest and are used widely in statistics or other subjects.With the development of society and technology,scholars are sparing no effort to study these two kinds of statistical methods so as to solve more realistic social problem and serve the development of the society.Mastering and using of model selection and model average methods can help us choose the accurate model.This paper make a continue study on some problems which related with model selection and model average methods that based on previous research result.This paper studied two issues in the respect of model selection method: one of the issue is that we find a new kind of cluster method when combing information criterion with cluster analysis in multivariate analysis;another issue is that we find a new method to chose a more accurate Poisson regression model by using information criterion function,at last in this problem we make an example to demonstrate how to use this method when we need to chose a best Poisson regression model.In the respect of model average method,we make an innovation of the average model's weight based on information criterion which produces a new way of model average method,then we make an average model and use it to do prediction to show this method's validity.