A Method of Grouping Projects for Comparisons

Most methods currently available for grouping an agency fundedprojects are well suited for continuous data (e.g., Dalenius and Hodges Cum frule) but not for grouping a few projects based on their means. This is because these projectmeans form a set of discrete observations. So, applying Dalenius and Hodges Cum frule to such discrete observations generally does not help in forming as homogeneousgroups as one would desire. Seeking more homogeneous groups in practice is necessary becauseoften an administrator needs as accurate comparisons among projects as possible (e.g.,evaluation of project performances) to make correct decisions about continuation or administrationof agency funded projects. Therefore, in this paper an iterative procedure is given togroup projects in such a situation. To apply this procedure, the only requirement is that the variableused for stratification is not a categorical or a nominal variable. The iterative procedure isillustrated by three examples. As should be expected, the procedure yields a different and morehomogeneous set of strata than the one obtained by the Cum f rule. For administration ofprojects, to form as homogeneous a group of projects as possible is important and , therefore,it is advisable to use this procedure to achieve more homogeneous strata for the data at hand.