科研成果详情

In the mean shift regression, it is of interest to detect anomalous observations that provide some large residuals or exert some unduly large influences on the least square analysis when the chosen model is fitted to the data, which are known as outliers or influential observations, respectively. The existence of outliers and influential observations, however, are complicated by the presence of a collinearity, which has great effects on the influences of a set of observations. In this paper, we show that when a ridge mean shift regression is used to mitigate the effects of the collinearity, the influences of some observations can be drastically modified. This is illustrated with an example derived from a set of data given by Mickey, Dunn and Clark[1]. Recommendations are given for obtaining the best use of the procedures.