QUICLSTR: A FORTRAN program for hierarchical cluster analysis with large numbers of subjects

Abstract

QUICLSTR is a program for hierarchical cluster analysis. It IS especially useful for problems involving large numbers of subjects. It can handle most such problems efficiently even on a fairly small computer, and does not require auxiliary storage of data during execution (cf. Levinson & Funk, 1973). The program uses squared Euclidean distance as the measure of dissimilarity between groups, and clusters according to Ward's (1963) error sum of squares method. At the user's option, data will be converted to standard (z) scores prior to clustering. Input to the program consists of (I) program parameters, read from card. and (2) raw data, read from card. tape, or disk. Output includes, for each step in the clustering process, (l) the error sum of squares, (2) the increase in this error over the previous step, and (3) thc acceleration of the increase of error. In addition, for as many steps as the user requests, the members of each group are listed. The primary limitation of the program is that when the number of subjects clustered is large, the number of variables employed must be correspondingly small. This limitation can