Variable Selection In Functional Regression Model

Generally,the object of multivariate data analysis in a research is to portray a number of observational data which presented in a number of statistical indexes.Sample data has the characteristics of discreteness and finiteness.However,the information which is collected by the modern data collection techniques,not only contains the data which is disposed by the traditional way,but also contains the data which is produced by the functional form process.Therefore,we will encounter the problem of founding model when we dispose the data.At this time we will apply multivariate data analysis model to functional data(for example,linear model).So the problem of how to select variables in liner model is important.During the process of analyzing this model,people always bring some dependent variable into the model.These useless dependent variables will result in large amount of calculation,and it will influence the precision of estimation and forecasting.Moreover,some observational data is very expensive.So this article will research how to select dependent variables in the process of analyzing functional data model.In this article,we will use the method which called lasso(Tibshirani(1996)).The main work is to research how to contract coefficients in the functional regress model and make some coefficients turn into 0.Then,using other way to bite off the coefficients which is 0.Through this way we can ascertain the order of the model,then we can select coefficients.