ABSTRACT
This chapter explains the use of statistical methods to analyse data from primary studies that have measured the outcome of two different treatments (also referred to as interventions). These could be the results of randomised experiments, or more general correlation or data mining studies. In the context of software engineering, a treatment would be a method, algorithm, process or technique for performing a software engineering process, activity or task. One of the treatments would be referred to as a control, if it corresponded
to effectiveness, often in terms of time (duration), effort, or faults. The aim is to calculate an overall summary measure of the comparative impact of the treatments based on the data from all relevant primary studies together with an assessment of its precision.
