一、引言在林业工作中,经常应用回归分析方法进行数据统计分析。然而,实际生产中所进行的回归分析一般是多变量的,如何从这多个变量中挑选出作用大的变量建立较好的回归模型呢?最佳模型应该是包含所有的有用变量,但又不多含一个无用变量,因为遗漏有用变量和多含了无用变量,对模型的预测预报都是无利的。解决这一问题的方法通常是逐步回归,通过对变量逐个筛选,选取对回归贡献较大的那些变量建立回归模型。但是在某一显著性水平 I. INTRODUCTION In the forestry work, we often use regression analysis method to carry out statistical analysis of data. However, regression analysis in actual production is generally multivariate. How to choose a large variable from these multiple variables to establish a good regression model? The best model should contain all the useful variables, but Not contain a lot of useless variables, because the omission of useful variables and contains unnecessary variables, the model of the forecast are unfavorable. The solution to this problem is usually a step-by-step regression, which establishes a regression model by selecting variables one by one and selecting those variables that contribute more to the regression. But at a certain level of significance