Beyond bar charts: A practical guide to small sample pharmacological data

In pharmacology and experimental biology, datasets are frequently limited in size (n < 20) due to the high costs and ethical considerations of data acquisition. In these "small-n" scenarios, estimating population parameters is challenging because single observations can disproportionately influence results. This creates a dilemma: while identifying anomalous observations is essential for reliable interpretation, ad-hoc data removal risks introducing investigator bias. Furthermore, many traditional outlier detection methods are optimized for large datasets and perform poorly at sample sizes typical of bench science. This manuscript provides a structured, logical workflow for exploratory data analysis (EDA) and objective outlier management tailored for the biological researcher. We introduce a "Visual Trio" framework incorporating Tukey boxplots, overlaid scatter plots, and "Confidence Funnels." This framework is used to assess data distribution and identify potential anomalies. We subsequently demonstrate the application of Dixon's Q-test as an objective diagnostic screen to guide sensitivity analyses in small samples. Furthermore, we address the special case of censored assay non-detects through prespecified sensitivity analyses rather than post-hoc exclusion. To support transparency and reproducibility, we provide annotated R code and links to public web resources, enabling scientists to maximize the value of their experimental data while maintaining statistical integrity prior to formal prespecified confirmatory studies.