自己顶一顶

For Univariate GLM, You have to compute it yourself.  For AIC, you need N, k (the number of parameters fit by the model, including the intercept), and RSS (the residual sum of squares).  For BIC, things are a bit more complicated; so unless you really want it for some reason, I'd stick with AIC (or the second order corrected version for smaller samples).

   http://en.wikipedia.org/wiki/Akaike_information_criterion
   http://en.wikipedia.org/wiki/Bayesian_information_criterion

This reminds me of something I've been meaning to suggest to the good folks at SPSS:  I think that AIC (and corrected AIC) would be very nice additions to the CURVEFIT procedure.   

Bruce Weaver Professor, Lakehead University Canada

SPSS中只能看残差平方和，你需要根据样本量，自变量个数和残差平方和自己计算出AIC\BIC，可以用excel自己搞一个公式
AIC=nln(残差平方和)+2（p+1）-nln(n)
BIC=nln(残差平方和)+（p+1）ln(n)-nln(n)
n是样本量，p是自变量个数