Scientific investigations into the safety of lithium-ion batteries often rely on small sample sizes, potentially leading to misinterpretations of test results. It is crucial to determine the optimal number of cells to test for obtaining meaningful and reliable results. This study focuses on 18650-type cylindrical cells with a capacity of 2.8 Ah and a NMC811/Gr chemistry, employing thermal stability tests to evaluate battery behavior. A total of 20 cells are tested, and five key parameters are identified and compared, with mean values calculated for each parameter serving as target values in the statistical analysis. Bootstrap sampling is conducted and random samples are taken from the measured values of the thermal stability test, with sample sizes ranging from 2 to 15. A total of 184,756 random samples are taken to explore all possible combinations when investigating sample sizes between 2 and 15 from the 20 measured values. For each sample with sample size k, mean values are calculated, followed by the calculation of the absolute difference between the sample size mean value and the mean value of the 20 cells. Further the probability is calculated, with which the mean value difference is smaller, than two defined accuracy criterias. The study demonstrates that parameters preceding thermal runaway can be accurately approximated even with small sample sizes. However, parameters directly linked to thermal runaway, such as maximum surface temperature and mass loss, exhibit significant deviations, necessitating larger sample sizes (greater than 15) for an accurate approximation of the mean value for 20 cells. Overall, this research highlights the importance of optimizing sample sizes in safety investigations of lithium-ion batteries to ensure the reliability and validity of the findings. By determining the appropriate sample size, researchers can enhance the accuracy of their assessments and contribute to the development of safer battery technologies.