Excerpt

There is ambiguity in the way medical publications indicate variation in bar graphs. Some of them present a short vertical line that depicts the magnitude of the standard error of the mean to facilitate visual assessment of the significance between two (or more) mean values. These error bars may be misleading because they indicate only the 67 % confidence limits [1] and because they are difficult to interpret without allowing for sample sizes [3]. In addition, they may give a false sense of individual variation [8]. Other publications display the standard deviation as a longer vertical line; this does indicate individual variability, and a standard deviation similar in magnitude to the mean suggests rightward skewing. Some figures even display confidence limits. At times, more than one of these displays occurs in different figures in the same publication, and often the metric chosen is not described [4, 7]. Many journals now require standard deviation to be used [2, 5, 6]. …