2.7. Plots for Detection of Multivariate Outliers

Sometimes a set of observations may violate certain model assumptions (e.g., data follows multivariate normal distribution). These observations are called outliers. The plots explained in the previous section can be used to detect possible outliers in the multivariate data. If one or more points fall outside the majority of the points on the Q-Q plot, then those points are suspected to be outliers. However, it is known that the statistics and S are both sensitive to the presence of outliers. Hence the squared Mahalanobis distance di2 calculated using the formula di2 = (yi - )′S−1(yi - ) may not ...

Get APPLIED MULTIVARIATE STATISTICS: WITH SAS® SOFTWARE now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.