What is the recommended method if 20 x < 40 with Gaussian distribution?

• A. Parametric with 90% CI of reference limits - must also include the histogram, mean or median, and minimum and maximum; alternatively, a table of all reference values can be provided along with the histogram
• B. Robust with 90% CI of reference limits
• C. Nonparametric with 90% CI of reference limits
• D. Bootstrap with 90% CI

Answer: (A)

When the data are roughly normal and the sample size is small, using a parametric approach is the most efficient way to estimate reference limits and their precision. The recommended method here is parametric with a 90% confidence interval for the reference limits because the normal model lets you derive the limits from the mean and standard deviation and quantify how precisely those limits are known.

In practice, you’d compute the central reference interval from a Gaussian assumption (for example, using mean ± a critical z-value corresponding to the 90% interval) and then provide a 90% CI around those limits to reflect uncertainty due to the small sample. Reporting should also include a histogram to show the distribution and either the mean with the min and max (or the median) or a table of all reference values, with the histogram, so readers can verify the normality assumption and the data spread.

Robust, nonparametric, or bootstrap methods are more appropriate when the data do not meet normality or when larger samples are available to stabilize percentile estimates; with a small Gaussian dataset, the parametric approach gives the most precise, straightforward estimation and clarity of the limits.