Building and Evaluating Single Variable Models
Alain Abran · 2015
This chapter explores how single variable models are built from an engineering perspective that is, based on observation of past projects. It shows how to build models with examples of data from the International Software Benchmarking Standards Group (ISBSG) repository, taking into account some of the peculiarities of software engineering datasets. The chapter identifies subsets within the ISBSG repository with enough projects for a statistical analysis according to their specific programming language. This chapter, illustrates an engineering approach to building single-variable models based on analysis of data from completed projects, analysis of the relationship to effort, one variable at a time, segmentation of a large dataset into meaningful samples relevant for estimation purposes, that is: excluding significant outliers and creating subsets of data taking into account the number of data points within size intervals. In engineering approach, the objective is not a single productivity model, but models that provide information.