Tutorial: Statistics for Human Biologists - Terminology and Univariate Statistics

Abstract

In the second part of this tutorial series, we will discuss the foundations of univariate statistics. We will define statistical terms such as p-value, confidence interval, and effect size. The latter two are often neglected, but they are more informative than the commonly overinterpreted p-value.

A p-value expresses the likelihood of observing a difference, or a larger, between a sample and a population parameter, or between two groups, assuming they share the same distribution. In univariate statistics, this difference is between a sample and a known population value. The assumption of equal distributions is the null hypothesis, which may be rejected if the p-value is very small. Importantly, the p-value depends not only on the difference itself but also on the sample size. Large samples can yield very small, highly significant p-values even when the effect size is trivial.

In contrast, a confidence interval provides a range for the population parameter based on the sample, giving insight into precision beyond statistical significance. The effect size conveys the magnitude of an effect, which simplifies interpretation and allows to assess the relevance of observed differences.

In this review, we explain these terms in the context of univariate statistics, comparing variables of interest from sample data against a known population value such as a mean or proportion. Relevant measures and visualizations include the mean, standard deviation, coefficient of variation, and histograms, barplots, along with effect size measures like Cohen’s d for numerical data and Cohen’s w for categorical data.