The validation by measurement theory of proposed object-oriented software metrics

Moving software development into the engineering arena requires controllability, and to control a process, it must be measurable. Measuring the process does no good if the product is not also measured, i.e., being the best at producing an inferior product does not define a quality process. Also, not every number extracted from software development is a valid measurement. A valid measurement only results when we are able to verify that the number is representative of the attribute that we wish to measure. Many proposed software metrics are used by practitioners without these metrics ever having been validated, leading to costly but often useless calculations. Several researchers have bemoaned the lack of scientific precision in much of the published software measurement work and have called for validation of software metrics by measurement theory. This dissertation applies measurement theory to validate fifty proposed object-oriented software metrics (Li and Henry, 1993; Chidamber and Kemerrer, 1994; Lorenz and Kidd, 1994).