Statistical analysis and prior distributions of significant software estimation factors based on ISBSG release 10

Abstract

Software estimation is an active research area with researchers working on areas like accuracy, new model development and statistical analysis. Estimates are probabilistic values and can be represented with a degree of uncertainty. Prior distributions are one of the way to represent the historical and organizational data which can be used by researchers to conduct further estimations. In this paper we introduce the software estimation landscape and prior distributions of significant factors, determined from ISBSG data set. These priors can be used for development of estimation models e.g. Bayesian networks. The paper make contributions in number of ways, it provides a brief overview of quality of data set. It also provides statistics of vital factors from dataset. This paper also provides prior distributions of productivity for Architecture e.g. Standalone, Client server and mixture of architectures. © 2011 Springer-Verlag.