Published at STAST'2020: Statistical Reliability of 10 Years of CS User Studies

This work continues our inquiry on meta-research in evidence-based methods in the Usable Security WP. In this work, accepted for the 10th International Workshop on Socio-Technical Aspects in Security (STAST'2020), we take aim at the reliability of statistical results from multiple angles. While our earlier work considered statistical reporting fidelity, our new work evaluates statistical tests computed in cyber security user studies for statistical power against effect size thresholds, publication bias and the winner's curse. Our work offers lessons learned and guidance for authors, reviewers and funders based on the empirical results.

Last modified: Thu, 26 Nov 2020 08:14:42 GMT