Abstract

Categorical data are central to clinical and epidemiological research, and appropriate statistical analysis is essential for drawing defensible conclusions. The one-sample chi-square test, also called the goodness-of-fit test, is used to assess whether observed frequencies are compatible with a specified expected distribution, such as equal group proportions, established genetic ratios, or a benchmark derived from previous research. Despite its broad use, the test is sometimes applied to inappropriate data or without adequate attention to its assumptions. This paper reviews common errors in health research, including applying the test to continuous measurements, overlooking small expected counts, violating independence, confusing goodness-of-fit with tests of association, and interpreting p-values without considering effect size or substantive importance. A hypothetical clinical example illustrates dataset preparation, step-by-step execution in SPSS, reporting of results, and cautious interpretation. Throughout the paper, we emphasise that a p-value describes the compatibility of the observed data with the null model, conditional on the assumptions of the statistical procedure; it does not represent the probability that the null hypothesis is true or that an observed result is due to chance.

Keywords

goodness-of-fit test, biomedical research, expected frequencies, statistical assumptions, independence, statistical significance, p-value interpretation