## Covratio

*Covratio* measures the effect of observations
on the covariance matrix of the parameter estimates.
For linear models,

where **X**_{(i)} is the **X**
matrix without the *i*th observation.
Values of *C*_{i} near 1 indicate that the observation
has little effect on the precision of the estimates.
Observations with suggest a need for further investigation.

For generalized linear models,

where **W**_{(i)} is the **W**
matrix without the *i*th observation,
**W** = **W**_{o} when the full
Hessian is used, and **W** = **W**_{e}
when Fisher's scoring method is used.
The Covratio statistics are stored in variables
named **C_****yname** for each response variable,
where **yname** is the response variable name.

Copyright © 1999 by SAS Institute Inc., Cary, NC, USA. All rights reserved.