chose to show only a single chart at once. For infrequent users, we felt an extra step to change the sort was worth increased clarity. We also realized users had diverse screen resolutions that could cramp a more complex layout. With only one chart at a time, we were able to add a second level of hierarchy with a parent bar for a category (like DNA vaccines) and child bars for individual values (one of six individual DNA vaccines).

The Plot supports y, x, and color variables (color and shape are varied together). With only a y variable, data values are jittered horizontally next to a box plot of their y distribution. Adding a categorical x variable creates additional jittered data columns and box plots for each x value (Figure 2). An interval x variable makes a scatterplot (Figure 4). A time x variable spreads the points by the study protocol day their samples were collected (Figure 5). Studies usually define Day 0 as the first vaccination. Data can be directly brushed to show all points from the same subject and selected in either one or two dimensions to create filters. Interestingly, design sessions with users have shown Plot can be useful for understanding the nature of the available data in addition to hypothesis generation. Plotting variables can be the most efficient way to answer questions like, “which types of vaccine have been tested against antigen X?” and also leads to natural follow-up questions about performance. Below, we describe several other plot features that support prioritized tasks.