Robust Estimate of the Covariance Matrix

The PHREG Procedure

Robust Estimate of the Covariance Matrix

The robust variance of Lin and Wei (1989) is a sandwich estimate given by

$\hat{\bld{V}}^s=\bld{I}^{-1} (\bld{U}'\bld{U}) \bld{I}^{-1}$

where $\bld{I}$ is the information matrix evaluated at the maximum likelihood estimate $\hat{{\beta}}$ and $\bld{U}$ is the n×p matrix of score residuals. Since the matrix of DFBETA residuals $\bld{D}$ can be written as $\bld{D}=\bld{U}\bld{I}^{-1}$ ,the robust sandwich variance estimate $\hat{\bld{V}}^s$ can be computed as

$\hat{\bld{V}}^s=\bld{D}'\bld{D}$

Chapter Contents
Previous
Next
Top