基于Lasso方法的平衡纵向数据模型变量选择

摘要

The Lasso method is applied to study variable selection problem in balanced longitudinal data model. This method can shrink the coefficients toward to zeros, and even set some coefficients to zeros, then LARS algorithm is used to sequence the coefficients, and AIC and BIC criteria are used to select the tuning parameters. Furthermore , some theoretical properties are proved, and the characteristics of the approach are presented from some simulation results. As an application, this approach is applied to find out the main factors which have influence to the activities of bats effectively.%应用Lasso方法研究平衡纵向数据模型的变量选择问题.通过Lasso方法可将模型的系数进行压缩并使之趋于零,甚至使一些系数等于零,利用LARS算法对回归系数进行排序,并采用AIC和BIC准则进行截取,从而达到变量选择的目的.同时证明该方法的一些理论特性,并从仿真模拟中分析了该方法的主要特点.作为实际应用,本方法可以有效地从众多的环境因素中寻找影响蝙蝠活动的主要因素.