Software Engineering Metrics and Models

by Samuel Daniel Conte

This text describes how software metrics and models can be used by managers, analysts, and programmers to develop more reliable and cost effective sofytware. It provides an introduction to software complexity metrics, the factors which affect software development productivity and cost estimation. It also discusses methodologies and statistical tools suitable for researchers to conduct their own experimental of proposed metrics or models.

Contents

1. The role of metrics and models in software development
2. Software metrics
3. Measurement and analysis
4. Small scale experiments, micro-models of efforts and programming techniques
5. Macrro-models of productivity
6. Macro-models for effort estimation
7. Defect models
8. The future of software engineering metrics and models

References
Appendix A
Appendix B
Index

396 pages, Hardcover

First published March 1, 1986

About the author

Samuel Daniel Conte, American computer scientist, educator. Served with Army of the United States, 1943-1946. Member American Mathematics Society, Association for Computing Machinery, Society for Industrial and Applied Mathematics, Mathematics Association American.

Reviews

[Alejandro Teruel · Rating 2 out of 5]

Good coverage of software complexity and software development metrics before 1985. Unfortunately most of the material has either fallen by the wayside or been superseded by more recent data or models. Take into account that Barry Boehm published the first version of his COCOMO model in 1981 (yes, it is covered by Conte, Dunsmore and Shen), before COCOMO II (2000), before agile and key object-oriented development.

Mainly of historical interest, although some aspects of chapter 3 (Measurement and Analysis) which covers material on the difficulty of gathering data, data transportability, the aging of data, controlled experiments (the validity of experiments, pre-experimental design, pre-tests and post-tests, counter-balanced design), statistical analysis, model evaluation criteria (coefficient of multiple determination, relative error and mean relative error, prediction at level 1, and the mean squared error and the relative root mean squared error criteria) probably still remains relevant.