DELETION DIAGNOSTICS FOR TRANSFORMATIONS OF TIME-SERIES

Deletion diagnostics are derived for the effect of individual observat ions on the estimated transformation of a time series. The paper uses the modified power transformation of Box and Cox to provide a parametr ic family of transformations. Inference about the transformation param eter is made through regression on a constructed variable. The effect of deletion of observations on residuals and on the estimate of the re gression parameter are obtained. Index plots of the diagnostic quantit ies are shown to be highly informative. Structural time series modelli ng is used, so that the results readily extend to inference about regr ession on other explanatory variables.