Dear All
I have a question concerning the Sandwich estimates of standard errors
commonly used in survey methodology when analysing clustered data. For the
model given by
y = XB + e
I understand that sandwich estimates of the standard errors do not require
that the residuals are homoscedastic or that the variance of these residuals
are homogeneous between clusters. My question is,do we assume that the
residuals are normally distributed ?
Many Thanks
Trevor Mole
Department of Medical Statistics and Evaluation
Imperial College School of Medicine
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|