1. Introduction

The study and measurement of software quality has contributed to the advancement of software engineering by providing ways of quantifying software systems which can lead to objective management decision making processes. Prior work has shown that a usable relationship exists between software measures and software quality.1 ^,2 ^,3 ^,4 Models exhibiting high levels of predictive quality can exert some measurable influence on the overall quality of a software system.

It is important to remember that the results obtained from such research are often difficult to apply in environments other than the one in which they were originally developed. So we caution the practitioner to be acutely aware when attempting to apply a model developed in one specific environment to another environment without validating it first. In cases where a model cannot be directly applied, we can still use the modeling methodology. The idea is to develop a model specific to the new environment provided that the model assumptions are not violated. For example, the data collected from software systems used to model software faults often violate the normality assumption of multiple linear regression (MLR) modeling. Applying such a model to this particular data set may not be a good choice. To analyze such data, researchers investigate other modeling methods whose assumptions, or...