The Impact of Interval Choice in Grouped Frequency Tables on Statistical Modelling

This paper examines the adequacy of grouped (interval) frequency tables for statistical modelling. Inspired by the chosen real-world data structure, the research question is: Can accurate modelling be achieved with the given grouping schemes without a significant loss of accuracy compared to the original data? To answer this, simulations based on log-normal distributions and various levels of grouping detail were conducted. The results show that large sample sizes enable accurate estimates even with low detailed censoring, provided the model aligns with the data-generating process. However, the mismatch between fitted and real distribution can introduce an additional bias, which can be reduced with detailed right-tail intervals. Therefore, it is recommended to consider this when choosing intervals for grouped frequency tables.