Pynnonen, Seppo (1992). Detection of outliers in regression analysis by information criteria. Proceedings of the University of Vaasa. Discussion Papers 146.

Abstract In this paper detection of outliers in the usual linear regression model by Akaike's and Schwarz's information criteria is considered. In terms of these criteria the problem can be considered as estimating the number of (dummy) variables in the model. By this method one does not have to concern the definition of underlying distribution of the observed residuals, $\hat\epsilon_j$, which in practice has proved to be very complicated (see e.g.\ Barnett and Lewis 1984, ch. 10). The method is illustrated by analyzing some well known data sets