In order to build a principal component classifier, which will flag those events that deviate from the normal connections, we need to calculate the distance between a record and the distributions of normal connections. We are going to use a distance metric, the **Mahalanobis distance**, which measures the distance between a point and a distribution. For the standardized principal components, like those here, the equation to compute the **Mahalanobis distance** is as follows:

*C _{i}* in this equation represents the value of each principal component, and

*var*represents the variance of each principal component. Let's ...

_{i}