教科书上是这要说的:
Forward: starts with the best single regressor, then finds the best one to add to what exists, then the next best, and go on.
Stepwise: simlilar to FORWARD, excpet that there is an additional step where all variables in each equation are checked again to see if they remain significant after the new variable has been entered.
假设有 independent variables x1, x2, x3, forward 会依据adj-R^2/R^2来选择初始的方程,然后依次加入方程,直到新加入的variable并不能通过测试(记得老师是这样说过的)
而stepwise就会增加检验新增variable的方程的level of significant.
就个人体验而言,SAS并不能十分的体验出两种方法的太大差别,
建议如果有spss的话可以用spss来体验一下区别!
spss中可以明显感受到stepwise在建模方面比较好,可以反复检验增加或者删减variable后的模型!
从而达到adj-R^2比较大的模型!(也许SAS也有这种功能,希望各位可以指出)

谢谢大吓...有点明白....

我不懂SPSS...