Biostatistics errors in a thesis are more common than most researchers admit — and more consequential than many realise. A viva examiner who spots a fundamental statistical mistake will not simply note it as a limitation. They will ask you to justify it, and if you cannot, you may face major corrections or, in serious cases, a failed viva.
The most common biostatistics mistake is using the wrong test for the data type. Running a Pearson correlation on ordinal Likert-scale data, or using an independent samples t-test on paired observations, invalidates the statistical inference even if the p-value looks clean. Always identify your measurement level — nominal, ordinal, interval, or ratio — before selecting a test.
A second frequent mistake is ignoring the assumptions of the test chosen. Every parametric test has assumptions: normality, homoscedasticity, independence. Running a one-way ANOVA without checking Levene’s test for equality of variances, or running a linear regression without inspecting residual plots, is incomplete analysis. Report your assumption checks in the methodology chapter.
Misinterpreting p-values is pervasive. A p-value of 0.04 does not mean there is a 96% chance your hypothesis is true. It means that, if the null hypothesis were true, you would observe results as extreme as yours 4% of the time. This is not the same thing. Report effect sizes (Cohen’s d, eta squared, r, R²) alongside p-values to give readers a sense of practical significance.
Multiple comparisons inflation is a fourth common error. If you run 20 independent hypothesis tests at α = 0.05, you expect one false positive by chance alone. Apply Bonferroni correction, Benjamini-Hochberg FDR adjustment, or — better yet — plan your primary analysis carefully so you are not running 20 post-hoc comparisons without a clear hypothesis for each.
