ABSTRACT
The use of multiple–partial correlation coefficients allows the researcher to make ample use of multiple indicators while at the same time retaining a manageable number of predictions. A common problem encountered in testing complex causal models is the selection of indicators. A common solution is the selection of one indicator that seems "on its face" to best represent the construct we wish to measure. This involves a loss of information and accuracy, as the two or three indicators that are thrown out are likely to have some validity, and their addition may produce a more correct representation of the construct. A second solution is that of combining the three or four indicators into an index in an attempt to represent the construct more accurately than any one of the indicators could possibly do.
