Choose categories that are mutually exclusive, exhaustive, contextually meaningful, and broad enough for every expected frequency to be greater than . If expected frequencies are too small, combine logically adjacent categories before performing the chi-squared test.
This is HL-only content from AHL 4.12: Data collection, reliability and validity tests. The explanation is paper-neutral because the same categorization principles apply whenever this topic appears on an HL paper.
How to Choose the Categories
Each observation must fit into exactly one category, while every possible observation must be included. For numerical data, use clear, non-overlapping intervals such as and , rather than ambiguous labels such as “low” and “medium”.