What is the recommended method if 40 ≤ n < 120 with Non-Gaussian distribution?

• A. Robust with 90% CI of reference limits (preferred)
• B. Parametric with 90% CI of reference limits
• C. Nonparametric* (Cannot determine 90% CI nonparametrically with < 120 reference samples; alternative methods, e.g., bootstrap, required)
• D. Bootstrap only

Answer: (A)

When the data are not normally distributed and the sample size sits between 40 and 120, using a robust method to estimate the reference interval with a 90% confidence interval is the best choice. Non-Gaussian data can be skewed and contain outliers, so methods that assume normality (parametric) can misestimate the true limits. A robust approach uses statistical estimators that are not unduly affected by skew or outliers, yielding a reliable central reference range and a 90% CI around the lower and upper reference limits even with a modest sample size.

Nonparametric CI around the limits relies on accurately estimating the tails of the distribution, which needs about 120 observations to be stable; with fewer samples, the CI becomes unreliable. Bootstrap can be an option, but the recommended strategy in this size range is the robust method because it provides stable reference limits and confidence intervals without assuming normality. Parametric methods are less appropriate here due to the non-Gaussian distribution.