Bad Chart Thursday: Scholarly Journal Edition
This week’s bad chart comes from none other than the prestigious peer-reviewed journal Cell, in an article titled “Regulators of Gut Motility Revealed by a Gnotobiotic Model of Diet-Microbiome Interactions Related to Travel.”
As you can see, the study involved mice playing a game of dietary musical chairs, followed by mincing a Bangladeshi child with a screen, then feeding the bits to the wild mice. I’m not positive, but I think that “turmeric” is a euphemism. If so, the study demonstrates that sprinkling powdered baby on your food may increase gut motility. Promising research for those who are sick of laxatives and happen to have a few extra babies lying around.
The research involved transplanting into mice the fecal microbiota of humans from different geographical locations whose diets varied from each other, then feeding the mice each of the different diets in sequence.
Figure 1 in the paper is perhaps the very first peer-reviewed word cloud in scholarly journal history:
The figure looks like something you’d find in a fitness-tracking app, not a biology journal. It is intended to depict the six different diets used in the research. The size of the font shows the proportions of the foods in the diets based on ingredient weight. The pie chart shows the proportion of macromolecules–carbohydrates, protein, and fats–in each diet.
Keep in mind that these are the diets of the specific people used in the study, not a typical diet of people from each location. Or maybe I’ve just totally missed out on the U.S. obsession with lima beans.
The primal diet also comes from a person in the U.S., a diet that heavily reflects our primal ancestors’ cow, pig, and chicken farming ways (“Old MacDonald had a cave, ee i ee i o. . . .”)
Neither the word clouds nor the pie charts show any actual data. For that, according to the figure caption, you have to look at Table S1. (This table is available for free on cell.com. Click to see the standard view, then click on the Supp. Info. tab.)
This table shows the characteristics of the human donors and their diets. For example, we (kind of) see why they chose to show a Bangladeshi baby in the graphical abstract. Although the initial experiment used adults, the subsequent experiments used a 2-year-old Bangladeshi donor (so not exactly a baby, but I think it’s safe to surmise that the authors’ clipart package was limited). All other donors were at least 21 years old. The authors explain using a child donor rather than an adult as a way “to avoid the potential confounding effects of chronic antecedent turmeric exposure on microbiota features,” although I’m not sure why they wouldn’t use children and adult donors for each location in the study to control for confounding factors related to age.
But back to our bad chart: I was particularly interested in seeing the data behind the whole milk and lima beans, and sure enough, lima beans were the largest proportion of the unrestricted U.S. diet by weight. In the primal diet, however, carrots should have been the largest word, with the highest weight of 43.5g/100g, compared with whole milk at 14.5g/100g. Pork chops should also be larger than whole milk, weighing 18.1g/100g. This reveals a significant downside to not including the data in the charts. The numbers might have been a visual check, alerting the authors to a discrepancy. The other downside of not including any data, of course, is that the charts end up being rather useless.
The upshot is that not even peer-reviewed journal articles are immune to bad chart design. Word clouds are a particularly odd choice for an academic paper considering they were already falling out of favor for far less rigorous uses 10 years ago, when Jeffrey Zeldman referred to them as the “new mullet.” But I wouldn’t go so far as to say that I got no information from today’s bad chart. I now know exactly what to sprinkle on my lima beans.
H/T to Andrew Gelman at Statistical Modeling, Causal Inference, and Social Science