A number of
models and theories in information systems research include concepts
of a match between two variables or states. The development of
measures for this concept can present problems, because decisions
must be made about the nature of the comparison. Should
indirect measures of the match be employed, then methodological
issues arise about how to best handle the measure when testing the
model. Difference scores are commonly used to measure a match
between variables or states in IS research, but these have implicit
assumptions about the theory and data characteristics that are often
false. Not unexpectedly, false assumptions can lead to
erroneous conclusions about the relationships among the variables
that are used to determine a match in a research model. The
implicit assumptions restrict the form of the relationships and
limit the IS researchers' ability to understand the possible
interplay among theoretical concepts. We suggest some
guidelines for the formation and testing of models that measure the
match. In addition we recommend polynomial regression analysis
as one means of analyzing the more complex relationships in IS
studies. We then use an IS service quality example to
illustrate the issues involved in the use of matching variables and
make suggestions with regard to using or avoiding difference scores.
Keywords: Difference scores, indirect measures,
polynomial regression analysis, IS SERVQUAL