Tags

, , , , , , , , ,

Great scientific data presentation is like great art. In my experience, it takes a lot of time, patience and practice to become merely competent at it, let alone to master it.

I consider myself fortunate to have attended Edward Tufte‘s one-day course on presenting data and information. If possible to do so (if living in the US or easy to visit US), I would recommend attending that. Since it’s quite expensive, I recommend getting a sponsor/employer/mentor to pay for it (which is how I attended it). A bonus of attending his course is that one gets his beautiful, encyclopedic books for free (at least that was the case when I attended). They remain for me a great resource for maintaining or at the very least attempting to maintain good taste in the art of data visualization and presentation.

I would not use PowerPoint as the primary source for putting together the data. I think Edward Tufte rightfully disdains PowerPoint as a poor tool for conveying scientific information. When used with its default settings and layouts, PowerPoint dumbs down presentations and ideas. Given its unfortunate ubiquity as a presentation tool (though Prezi is beginning to catch on these days), it’s critical to understand its limitations as a tool for presenting a compelling narrative and indeed for conveying useful information. To better understand these limitations, I recommend looking at  Peter Norvig‘s PowerPoint version of Lincoln’s Gettysburg address The Gettysburg Powerpoint Presentation. It is not only amusing but also extremely effective.

I make the original plots in GraphPad Prism (Company – graphpad.com). Unlike many other charting programs, it was developed under the guidance of a scientist, Harvey Motulsky. I believe this is a key reason I have found it easy and intuitive to use for plotting scientific data. It is quite an expensive piece of software so having institutional access to it would be quite helpful. Prism plots can be directly sent to PowerPoint and Word with a simple click. This makes it quite easy to assemble a PowerPoint presentation of Prism data plots.

I usually spend a lot of time plotting a given piece of data using different types of plots: scatter (aligned or not), box and whiskers, before-after, x-y line, histogram (my least favorite choice; it hides a lot of sins [read variability]), etc. I do this because I have learned that simply visualizing the same data in a variety of ways helps me better understand it. However, doing so is also a double-edged sword: the better one gets at this, not only does the ability to understand the data improve but so does the ability to manipulate it.

In my experience, even the best-intentioned graphing software has within it the inherent tendency to add unnecessary Chartjunk. Therefore, regardless of the software one uses to plot the data, I suggest to keep in mind Edward Tufte‘s tips. I call this “Tufte-izing” the data plots. Some of the ones I use are:
1. In the graphs, reduce the amount of “non-data ink”. Usually, many default settings are such that the data plots have features that detract from the data. For example, thickness of axes, choice of font and font size, unnecessary lines that are part of the default design, etc.
2. Where appropriate, use “Small multiple” for visualizing data series. This is a way of using the same type of data plot over and over again. The way we process visual information, “Small multiple” help to make differences between similar data sets pop out.
3. Whenever possible, use color to convey information. As Edward Tufte often points out, this is very much an under-utilized tool.

The key to a good data presentation is to master the story that the data is telling. This is why I spend a lot of time plotting the same data various ways to master its nuances as much as I can. Once that is done, the presentation tool is no longer center-stage, and indeed is much less relevant. Under such circumstances, I have found to my shock that even something like PowerPoint is tolerable.

What are the best ways/quick tips to present scientific data in PowerPoint?

Advertisements