Hi Class,
    We are trying to set priors when using Amelia for imputation, but it keeps return errors like "The resulting variance matrix was not invertible.  Please check your data for highly collinear variables"

The priors look something like:
pr<-matrix(NA,6,5)
for (i in 1:6){
   pr[i,1]<-i+48
   pr[i,2]<-47
   pr[i,3]<-1.81
   pr[i,4]<-2.01
   pr[i,5]<-0.9
}  


Amelia runs without any problem if we don't set priors. Does anybody know what the problem might be?


Many thanks,
Yue




On Sun, Apr 24, 2011 at 9:36 PM, Leslie Finger <lfinger@fas.harvard.edu> wrote:


Dear Class,

My partner and I are clustering the standard errors in new way (that we think
makes more sense in theory than what the authors were doing), but now, while
some clustered standard errors are very slightly bigger than the lm output,
some standard errors are actually smaller (in the same regression).  Does
anybody have any intuition about what this might mean?

-Leslie and Adela

_______________________________________________
gov2001-l mailing list
gov2001-l@lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l