Highlighted Selections from:

Data visualization: ambiguity as a fellow traveler

DOI: 10.1038/nmeth.2530

Marx, Vivien. “Data Visualization: Ambiguity as a Fellow Traveler.” Nature Publishing Group 10.7 (2013): 613–615. Web.

p.613: Data from an experiment may appear rock solid. Upon further examination, the data may morph into something much less firm. A knee-jerk reaction to this conundrum may be to try and hide uncertain scientific results, which are unloved fellow travelers of science. -- Highlighted mar 11, 2014

p.613: Statistical uncertainty weighs heavily on visualization. Every data point has uncertainty associated with it, Krzywinski says. Adding those statistical data to visualizations can quickly overload them. Despite the potential pitfalls of including uncertainty, the visual cues can remind scientists of their data's ambiguity. -- Highlighted mar 11, 2014

p.613: Uncertainty comes in many flavors. It can arise upon data capture, during analysis or during visualization. It may be due to missing, noisy or imprecise data or to filters that could skew calculations, or there may be too few data to begin with, says Heidrun Schumann, a computer scientist at the University of Rostock who studies uncertainty visualization in many research areas, including the life sciences. -- Highlighted mar 11, 2014

p.614: One of the most challenging facets of uncertainty for scientists is visualizing which data are or may be missing. -- Highlighted mar 11, 2014

p.614: With large data sets that have millions of data points to be visualized, values pile on data points to be visualized, values pile on top of others, erasing or hiding one another. The lost data become visible only when examined at high resolution. Visual indicators tell the scientists to look more closely at a row “to see the information that had gone missing,” she says. Without this flag, a researcher would have no cues for which data to examine more closely. -- Highlighted mar 11, 2014

p.614: Both scientists are exploring ways to build uncertainty visualization into their platforms and, in particular, ways to highlight for scientists what might be missing in their data and their analysis steps. Information is often missing about where data originated and how they were processed, says Gehlenborg, who was interviewed jointly with Lex. “We keep track of where the data come from,” chronicling the analysis steps and retaining the metadata, too, he says. His idea for Refinery is to allow scientists to toggle through their data and analysis steps. The software visually heightens awareness of what is missing at a given step -- Highlighted mar 11, 2014

p.614: Refinery will keep track of analyses performed on a data set. Without such tracking, uncertainty information can get buried in analysis steps that cannot easily be teased apart, he says. -- Highlighted mar 11, 2014

p.615: It is all too common in labs to work with the summaries and the results, without a way to return to the raw data and all of their variability, which have scientific value. -- Highlighted mar 11, 2014

p.615: Visualization methods have to keep up with large data sets that are big, complex and noisy, but they cannot replace statistics. Visualization partnered with statistics stands to become a powerful part of biology-related data analysis, says Gehlenborg. -- Highlighted mar 11, 2014