Reproducibility of Statistical Tests Based on Randomised Response Data

Reproducibility of experimental conclusions is an important topic in various fields, including social studies. The lack of reproducibility in research results not only limits scientific progress, but also wastes time, resources, and undermines society’s confidence in scientific findings. This paper focuses on the statistical reproducibility of hypothesis test outcomes based on data collected using randomised response techniques (RRT). Nonparametric predictive inference (NPI) is used to quantify reproducibility, which is well-suited to treat reproducibility as a prediction problem. NPI relies on few model assumptions and provides lower and upper bounds for reproducibility probabilities. This paper concludes that less variability in the reported responses of RRT methods leads to higher reproducibility of statistical hypothesis tests based on RRT data with the same degree of privacy.