Graphical Tools for Visualization of Missing Data in Large Longitudinal Phenomena

Abstract

The analysis of large quantities of longitudinal data requires quick decision tools to ensure data quality and to find useful patterns for analysis in exploratory stages. We propose algorithms based on ordering, sampling and grouping applied to lasagna plots, a special kind of matrix plot, which are heat maps created to visualize longitudinal studies. These algorithms can be applied to large data sets to find patterns of interest, monotone and intermittent, in the missing data with low computational cost compared to previous alternatives. Visualization with these algorithms addresses a trade‐off in visualization design: reducing visual clutter versus increasing the information content in a visualization. The method enables the visualization of missing data in a clear and concise way. We apply our techniques to four real‐world data sets of different origins and sizes that share analysis and visualization tasks and discuss the patterns found within them.