Several techniques have been developed for detecting
problems with dataset in regression analysis. They differ in the
particular regression result on which the effect of a deletion of
an observation is measured. For instance, the Cook’s distance
measures the effect of observations on the estimated regression
coefficients. Other measures of influence measure the effect of
observations on the fitted values and the variance-covariance of
the parameter estimates. This paper is another attempt at
identifying a more responsive measure for detecting even the
more subtle suspect outliers