An empirical study of a model for program error prediction

A model is presented for estimating the number of errors remaining in a program at the beginning of the testing phase of development. The relationships between the errors occurring in a program and the various factors that affect software development, such as programmer skill, are statistically analyzed. The model is then derived using the factors significantly identified in the analysis. On the basis of data collected during the development of large-scale software systems, it is shown that factors such as frequency of program specification change, programmer skill, and volume of program design documentation are significant and that the model based on these factors is more reliable than conventional error prediction methods based on program size alone