Oregon Program Evaluators Network: Visual Displays of Data
One of the OPEN Annual Conference afternoon workshops was “Groovy Graphics: Visual displays of Quantitative Information” from Elizabeth O’Neill and Ralph Holcomb of Multnomah County Aging and Disability Services. (Silly presentation title, not clear why they used it, they didn’t seem to like it, either)
The presentation began with a “Graphics 101″ review of some of Edward Tufte’s concepts about using graphics that “concentrate on the eloquent statement”:
- Show comparisons;
- Provide data at different levels of detail;
- Integrate information so viewers can understand it quickly;
- Provide documentation;
- Reduce clutter.
Every visual choice should convey a meaning. MS Excel provides a toolbox of fundamentals for us:
- Bar Graphs–avoid using the 3D option because it serves no purpose and simply adds visual clutter; typical axis choices are to make the x-axis nominal and the y-axis ordinal, interval, or ratio; horizontal bar graphs are especially useful in facilitating comparison of data to a benchmark;
- Line Graphs–typically the x axis is ordinal or higher and the y axis is interval or ratio;
- Scatterplots–for continuous data; both x and y are interval or ratio;
- Stem and Leaf charts–show mean, median, and mode in one glance (but I find them a bit hard to read);
- Pie Charts–for displaying many values of one variable.
The presenter stated that he is a “strong advocate of DDT”–meaning Dashboards, Drill-downs, and Templates:
- Dashboards provide quick groups of data presented simply (often featuring traffic signals in green, yellow, and red);
- Drilldowns (web pages where viewers can click on a word or a tab) are ways to provide more data with less clutter;
- Trendlines show patterns of data over time.
Sparklines are tiny graphs, charts, traffic signals embedded within text, showing lots of data in a small space. They provide very quick views of data within the context of a narrative, and can be produced with an add-on that can be purchased for Excel (they will be a standard feature in the next version).
Word clouds, such as those produced from Wordle, can be used in textual data analysis for counting the occurrences of words. In word clouds, word size matters–bigger means more–but word position does not–it’s random. So, conclusions cannot be drawn from word proximity. But, the presenter suggested that word clouds could be given to decision makers to help them generate hypotheses and could be used in generating “frequently asked questions” lists. Word clouds also have good potential for use as graphics for the fronts of reports.
Handouts and powerpoints from this workshop will be made available at the OPEN web site.