A Quasi-experiment for Effort and Defect Estimation Using Least Square Linear Regression and Function Points

Software companies are currently investing large amounts of money in software process improvement initiatives in order to enhance their products' quality. These initiatives are based on software quality models, thus achieving products with guaranteed quality levels. In spite of the growing interest in the development of precise prediction models to estimate effort, cost, defects and other project's parameters, to develop a certain software product, a gap remains between the estimations generated and the corresponding data collected in the project's execution. This paper presents a quasi-experiment reporting the adoption of effort and defect estimation techniques in a large worldwide IT company. Our contributions are the lessons learned during (a) extraction and preparation of project historical data, (b) the use of estimation techniques on these data, and (c) the analysis of the results obtained. We believe such lessons can contribute to the improvement of the state-of-the-art in prediction models for software development.