Using blind analysis for software engineering experiments

Context: In recent years there has been growing concern about conflicting experimental results in empirical software engineering. This has been paralleled by awareness of how bias can impact research results. Objective: To explore the practicalities of blind analysis of experimental results to reduce bias. Method: We apply blind analysis to a real software engineering experiment that compares three feature weighting approaches with a naïve benchmark (sample mean) to the Finnish software effort data set. We use this experiment as an example to explore blind analysis as a method to reduce researcher bias. Results: Our experience shows that blinding can be a relatively straightforward procedure. We also highlight various statistical analysis decisions which ought not be guided by the hunt for statistical significance and show that results can be inverted merely through a seemingly inconsequential statistical nicety (i.e., the degree of trimming). Conclusion: Whilst there are minor challenges and some limits to the degree of blinding possible, blind analysis is a very practical and easy to implement method that supports more objective analysis of experimental results. Therefore we argue that blind analysis should be the norm for analysing software engineering experiments.