Five principles for the formal validation of models of software metrics
Lem O. Ejiogu · ACM SIGPLAN Notices · 1993
The goal of software metrics is the improvement of the software process. Five principles are fundamental to the formal validation of models of software metrics:-- Attribute Type --what attribute(s) of software behavior are we measuring?; The mathematical postulates of a measure function --a model must have the utility for comparison; this unique criterion must be passed by every effective model of measure; Metrical Vindication --applying the model directly to measure actual software module(s); how consistent and reliable are the results of the measures?; Feedback Effect --how does the metric help us to identify faults and errors (of design) or improve testing and maintenance; what is a metric if it is just a number?; Units of Measure --what are we measuring with respect to dimensions: size, length/depth, extent, degree of variation, degradation, etc.?; a measure without unit(s) is like a building without a roof.Without principles, there is nothing to validate. Every good science is a body of relevant principles. Some worked examples on some of these principles are presented. Some of these examples in a way, expose the failures or frustrations resulting from the applications of some models of measure presently being used in industry and science. Not every mathematical function can be used as a model of measurement.