Comparing Two Independent Proportions Using the Z-Test: Improving the Statistical Quality of Our Studies

Abstract

In health sciences research, it is common to compare percentages or proportions between groups without using appropriate statistical techniques to validate whether the differences are significant. This simplistic approach relies on the visual comparison of values ​​in tables or graphs, ignoring sampling variability and confidence intervals. As a result, statistically or clinically relevant differences are erroneously assumed, leading to incorrect conclusions and affecting decision-making. This error is frequent in basic epidemiological studies or publications with low methodological rigor.

Let's look at two examples below. Remember that the p-value measures the probability of observing differences, assuming that no real difference exists (null hypothesis). If p < 0.05, we reject the null hypothesis. The 95% confidence interval (CI) shows the plausible range of the real difference; if it does not include 0, there is evidence of a difference.