If outlier detection methods cannot be appropriately applied due to non-Gaussian data that cannot be transformed, which method is preferred for RI determination?

• A. Nonparametric methods are preferred for RI determination because they are less influenced by outliers
• B. Parametric methods are preferred for RI determination because they assume normality
• C. Bootstrap methods are preferred for RI determination
• D. Bayesian methods are preferred for RI determination

Answer: (A)

When data are non-Gaussian and cannot be transformed to meet normality, using methods that do not rely on distributional assumptions is essential. Nonparametric approaches for RI determination rely on the observed order of values rather than on the mean and standard deviation, making them robust to skew, heavy tails, and outliers. By defining reference limits using percentiles (for example, the central 95% of the data), you obtain a data-driven interval that faithfully reflects the actual distribution, without needing a normal model. This avoids the distortions that occur with mean±SD when the data are not normal, so the resulting reference interval is more reliable under these conditions.