Statistical Modelling of Software Metrics

The development of quality software products has been of critical importance to modern industry and commerce for some time. However, software is still generally regarded as being both unreliable and costly. One suggested solution to this problem is the application of scientific measurement and statistical principles.

Statistics and software measures (metrics) can be used in software engineering in a number of fundamental ways. These include scientific experimentation, certification and assessment and management of industrial software projects.

This dissertation proposes a framework for the effective application of statistical methods and software metrics. The framework identifies the aims, pit-falls and technical considerations needed by researchers and practitioners when applying statistics to software metrics data. Examples of
the implementation of this framework are extensively used to both illustrate the principles of the framework and to promote the use of statistics in a number of areas. Specific attention is given to defining the role of statistics
in software product certification, process management, data linkage and the validation of theories.