Ad Lagendijk
22 April 2008
Selling science with candy sticks and flying carpets
Posted in High-impact journals
Not only the scientific glossy magazines but also the professional journals are invaded by the colorful graphs that either look like an artificially colored moon landscape, or like a collection of candy sticks or like a flying carpet. I do not like them and in particular the candy sticks I find ugly.
In physics and mathematics progress is made by generalizing and abstracting. The ultimate result is the capturing of one’s finding in a mathematical formula or an – admittedly dull – scientific X-Y graph. As a result physics papers were are full of these terrifying items. I say were, as the X-Y plot is on its way out.
The 3D plots are usually very difficult to interpret quantitatively. They are good to get an overall picture. It gets even worse if the authors – often on purpose – do not show the color scale. Wildly varying colors sometimes only indicate a variation of the magnitude of a variable by less than 10%.
Scientists that compare experimental results with theoretical calculations by showing them both in the same X-Y plot are honest and have guts, as this type of plot shows disagreement without any mercy. If the comparison between calculation and observation is presented by using magic carpets it usually means that the authors want to hide the poor agreement.
But maybe you think these 3D-color plots add a lot to the quality of a scientific paper
and consider my observations as coming from a complaining old-fashioned guy.
Or subscribe to our RSS feed.
24 Apr 2008 9:52, Sanli
With the automation of experimental setups and data-acquisition processes, the amount of data that is acquired for some experiments is extremely enormous for some cases. There exist a point, which you may also agree with, that an example of the raw data must also be accessible to the readers of an article. What is your suggestion, other than 3d plots or color-maps, for presenting the data which does not fit a couple of X-Y plots? (Coulomb diamonds, enhanced backscattering spectrum, and energy landscapes are just a few examples)
8 May 2008 8:26, Patrick
For graphically representing the relationships among three numeric variables in two dimensions I recommend contour plots. These allow both data and fit to be presented. Small deviations between data and fit are immediately visible. Also, the value of each contour level can be precisely stated. Finally, they can be presented in black and white, so you don’t need a color printer.
13 May 2008 0:57, bill peppin
I’ve used my own graphics software for many years, and find that not many things are assisted by using 3-d plots. But in the image business, certainly some things are, e.g., in my biz, draping some sort of x-y info on top of a 3-d rendering of topography, or over a 3-d rendering of buildings in a city, can be the best visualization tool. But why stop at 3-d, when nowadays typical problems are located in spaces of some high dimension. My belief is that 2-d cross sections of such spaces provide the clearest, easiest-to-understand view, which is why, even having access to multidimensional plotting, the stuff I make regularly are 2-d. My view is that 3-d plots are mainly to be used in marketing, wherein the audience wants to be “wowed” by some presentation absent much, if any, understanding of the glitzy thing being plotted. For sure, much of what is presented falls into the category of chart junk and is, in honest presentation, to be avoided at all costs.