When this option is chosen, JMP weights each variable by the inverse of their variance before using the classical root sum of squares to compute the distances between points. Choose this option to use a full, pooled covariance matrix across all variables as an inverse weight matrix.