Application of Statistical Meta-Analysis to Software Engineering Metrics Data

摘要

This work reports an a study on the feasibility of applying "weighted estimator of a common correlation" technique to software measurement data. This technique is a meta-analytical approach to generalizing the knowledge from individual statistical studies. The technique is applied on a data set of 100 public domain Java projects to determine if it is possible to generalize project-specific significant correlations. The results of the experiment show that the technique avoids both type I and type II errors.