Supplementary survey analysis and MATLAB scripts for
Data visualization in the neurosciences: overcoming the curse of dimensionality
Elena A. Allen, Erik B. Erhardt, and Vince D. Calhoun
EA Allen, EB Erhardt, VD Calhoun. Data Visualization in the Neurosciences: Overcoming the Curse of Dimensionality, Neuron 74 (2012), doi:10.1016/j.neuron.2012.05.001


Abstract
In publications, presentations, and popular media, scientific results are predominantly communicated through graphs. But are these figures clear and honest, or misleading? We examine current practices in data visualization and discuss improvements, advocating design choices which reveal data rather than hide it.

Supplementary survey analysis and Matlab scripts

• 2012_Allen_DataVisualization_SupplementaryNote.pdf includes the neuroscience figure survey results and creating dual-coded images shown below.
• dualcodeExample.zip includes the Matlab script files and example data to "Show more, hide less!"

Neuroscience figure survey results

To formally assess the clarity and completeness of recently published graphics, we sampled 288 articles published in 2010 from 6 leading neuroscience journals (Frontiers in Systems Neuroscience, Human Brain Mapping, Journal of Neuroscience, Nature Neuroscience, Neuroimage, and Neuron) and examined the 1451 figures therein. See pdf.

Survey results separated by journal. (A) Mean proportion of 2D (white) and 3D (dark gray) figures displaying each feature. Error bars denote 95% non-parametric confidence intervals (10,000 resamples). NN = Nature Neuroscience; NE = Neuron; JN = Journal of Neuroscience; FSN = Frontiers in Systems Neuroscience; NI = Neuroimage; HBM = Human Brain Mapping. (B) Scatter plot of the mean 2D feature proportion (averaged over features 1-4) versus the mean 3D feature proportion (averaged over features 1-3) for each journal. Crosses indicate mean data values; gray ellipses denote 95% confidence regions.

Creating dual-coded images

To encourage the generation of rich displays for 3D data that portray uncertainty, we provide sample MATLAB scripts for hue and transparency coding. Scripts and summary statistics from an example fMRI auditory oddball (AOD) dataset (see Figure 3 of PAPERLINK) are included in dualcodeExample.zip. The script dualcodeImage.m will create a dual-coded image for a single axial slice, with quantitative effects mapped to color hue and effect uncertainty mapped to color transparency. See pdf.

dualcodeExample.zip includes the following Matlab script files and example data to "Show more, hide less!":

• dualcodeImage.m creates images with hue and alpha color-mapping with the following steps:
  1. Load the AOD_data.mat file with sample data from the fMRI AOD experiment,
  2. Set some defaults that will affect the appearance of the image (min/max values, labels for colorbar, colormap for unerlay and overlay),
  3. Do the actual plotting (plot underlay, alpha map based on t-statistics, set alpha for overlay, add contours), and
  4. Create a 2D colorbar for the dual-coded overlay.
• convert_to_RGB.m is used to transform pixel values with arbitrary scaling into truecolor RGB values.
• AOD_data.mat is the AOD_data.mat file with sample data from the fMRI AOD experiment.

To produce the image, make sure the contents of dualcodeExample.zip are on your path (or just un-zip and cd to the folder) and run dualcodeImage.m from the command line:

>> cd dualcodeExample
>> dualcodeImage

You should see something like this:

Result of dualcodeImage.

Please send comments and bug reports to vcalhoun@gsu.edu.