Jimmy passes this article by Ahmad Reza Hosseinpoor and Carla AbouZahr. I have little to say, except that (a) they seem to be making a reasonable point, and (b) those bar graphs are pretty ugly.
Jimmy passes this article by Ahmad Reza Hosseinpoor and Carla AbouZahr. I have little to say, except that (a) they seem to be making a reasonable point, and (b) those bar graphs are pretty ugly.
As a tech in an oncology lab, I am constantly plotting gene expression data (from PCR; polymerase chain reaction) using bar graphs similar to those horrendous graphs in the article. Two things that immediately come to mind:
1: Make the bars different colors
2: Align the "correct" image horizontally, with the baseline value of 1 representing 0 on the real line.
Other thoughts?
I find rel risks look much better plotted horizontally not vertically.
With CIs too.
If you use CIs you can maybe not use a log scale if they don't differ by to much and the customer insists.
DaveG
My problems with these bar graphs, which I indeed see them very often in the context of gene expression data, is that people use them to plot averaged data (often with a standard deviation) instead of actually showing the data. This way, they actually make sure that any problem in the data is hidden behind a largish standard deviation (e.g. one replicate very different from the other two or three replicates) instead of actually displaying interesting information !
As a side note, I did not know that it was possible to publish articles with so little information in the Lancet (not that their point is not reasonable, but it is seems a bit too obvious to require an article in itself).