J. Colihan et Gk. Burger, CONSTRUCTING JOB FAMILIES - AN ANALYSIS OF QUANTITATIVE TECHNIQUES USED FOR GROUPING JOBS, Personnel psychology, 48(3), 1995, pp. 563-586
A Monte Carlo study was conducted to examine the performance of severa
l quantitative grouping strategies for the purpose of grouping jobs in
to job families. Two factors were found to substantially affect the ac
curacy of these grouping strategies in terms of identifying the correc
t number of families, and accurately classifying jobs into those famil
ies. Through simulation of job analysis data sets designed to reflect
various underlying structures among a set of jobs, it was found that t
echniques based on the commonly used hierarchical cluster analysis mod
el were relatively inaccurate when applied to data containing measurem
ent error or overlap between job families. Alternatively, Q-type facto
r analysis and hybrid techniques involving a combination of factor and
cluster analysis proved to be viable and robust grouping strategies f
or job classification research.