在网上看到了这个 ，大家觉得是应该用固定效果还是随机效果呢？
hausman test correlated random effects                        

by kpcinthin on Mon Jul 19, 2010 1:57 pm

                                                After estimating my model with random effects specification, I run the correlated random effects -- hausman test, but results include the warning "cross-section test variance is invalid.  Hausman statistic set to zero."  Also, "robust standard errors may not be consistent with assumptions of Hausman test variance calculation."  Have I done something wrong in the initial estimation or how do I interpret the test results in light of these warnings?
                                       
                                                                        kpcinthin                         Posts: 1Joined: Mon Jul 19, 2010 1:48 pm                        Top
               
        
                                                                                                Re: hausman test correlated random effects                        

by EViews Glenn on Wed Jul 21, 2010 1:10 pm

                                                The Hausman test statistic estimator for the variance of the coefficient difference is not guaranteed to be positive definite. In cases where one cannot compute the statistic, we set the value equal to zero and report the first warning message that you encountered.

The robust message is generated whenever you are estimating using robust standard errors. In this setting, you are saying that you believe that the standard model is not efficient, which is a requirement for the computation of the Hausman test variance.

原文链接
http://forums.eviews.com/viewtopic.php?f=18&t=2688&p=9055&hilit=hausman+test#p9055