This is possible because, not restricted to the example you've given here, different package may use different convergence criterion. take another instance, for maximum likelihood estimator, different default stpe-length search method are used. For Excel, Minitab and spss13, the discrepancies is acceptable. However, I was also confused with the result from Eviews. Even if kurtosis in Eviews is not adjusted for normal distribution, there is still a drastic disagreement.
扫码加好友，拉您进群
全部版块
我的主页
收藏
