Re: E-cigarette use and associated changes in population smoking cessation: evidence from US current population surveys
Notwithstanding the merits of the article by Zhu et al, I noted with disappointment that their Figure in the printed version of the paper, and Figure 1 in the online version is one more example of poor graphical display and violates the so-called Tufte's principle [1, p6].of maximising the ratio amount of information conveyed to the amount of ink. The main issue is the use of some measure of uncertainty at the top of the bars in the bar chart. Nowhere is this defined and so we don't know what they are supposed to show. (eg standard error or confidence intervals). Secondary to this is that they do not measure uncertainty about the main outcome, which is the difference between the cessation rates for various groups. These confidence intervals are correctly given in the text, but cannot be derived directly from the figure without certain assumptions. One would also advise against the use of bar charts for displaying such data [1, p10] since the figure essentially only conveys six numbers and hides much information such as the sample size, which could be given in a table occupying the same space.
Freeman JV, Walters SJ. Campbell MJ, How to display data . Oxford: BMJ Books/Blackwell 2008.
Competing interests: I was a co-author of a BMJ book on displaying data