PHYSICS 102 Lecture Notes - Lecture 13: Linear Discriminant Analysis, Euclidean Distance, Distance Matrix
Document Summary
Discriminant analysis requires you to know group membership to derive the classification rule. Hierarchical clustering choose a statistic to quantify, select a method for forming the groups, determine how many clusters you need to represent your data: small data, easily examine solutions. K-means clustering select the number of clusters you want, algorithm estimates the cluster means and assigns each case to a cluster if you know how many clusters you want and moderate data size. Agglomerative: begins with every case being a cluster unto itself, algorithm ends with everybody in one useless cluster, once a cluster has been formed it cannot be split, only combined. Divisive: starts with everyone in one cluster and ends with everyone in individual clusters. Must select: criterion for determining similarity or distance between cases, criterion for determining which clusters are merged at successive steps, the number of clusters you need to represent your data.