Clustering gap-statistic. The plot is a display of the gap-statistic values (100 simulations for each K-value) for each choice of number of clusters from K=1 (all CGFs belong to the same cluster) up to K=10. The error bars under each circle is the standard error for each value. The optimal number of clusters is where we see the first significant drop in gap-statistic value, i.e. where the gap statistic value G(K) is such that G(K)>G(K + 1)−SE(K + 1), where SE(K + 1) is the standard error at K + 1.