Visualizations of Data Created to Aid in the Detection of Laboratory Data

Abstract

Research and service activities in fundamental and clinical pharmacology rely heavily on the assessment of pharmacological
and endogenous chemical concentrations. Using data science visualizations of laboratory data, it is shown on a real-life example
that recommended standard visual techniques of data inspection or basic statistical investigation of laboratory test findings
may fail to discover systemic laboratory mistakes. For instance, conventional data quality control approaches could miss data
pathologies like an experiment run where all probes provide the same result. We find that sys- tematic laboratory mistakes may
be better detected when using data visualizations that highlight diverse perspectives on the data. To better understand the data
range, outliers, and a specific kind of systematic error—where identical values are incorrectly obtained in all probes—a dotplot
of individual data organized by assay is suggested.