A Dynamic Framework for Determining Optimal Experimental Repetition Based on Data-Driven Criteria

Abstract

Experimental research plays a crucial role in the scientific method as a means to acquire reliable knowledge. Ensuring the dependability and reproducibility of an experiment requires careful consideration of the number of repetitions or runs conducted. This study proposes a dynamic and data-driven framework to determine the optimal number of experimental repetitions based on the specific requirements of each experiment. The proposed steps begin with a normality test, followed by statistical comparison across data sets, identification of the required number of runs, and finally post-hoc confirmation. Several statistical techniques were employed, including Kruskal-Wallis test, power analysis, Sequential Probability Ratio Test,  confidence interval estimation, and cumulative standard deviation to demonstrate the framework. Application of the framework to real-world electric field measurements obtained from a Giga-Hertz Transverse Electromagnetic (GTEM) cell demonstrates that the determination of an adequate number of runs is highly dependent on data behaviour and methodological assumptions. This open-ended yet structured approach empowers experimenters to design and analyze experiments with statistical rigor, enhancing both reliability and reproducibility.