Comments on: Tornado Charts and Dot Plots

By: Jon Peltier

Jon Peltier — Wed, 22 Jul 2009 04:19:39 +0000

Here are a couple links: Tornado Charts Tornado Charts and Dot Plots Keep in mind that tornado charts are not as good as you might think for comparing two sets of data. The fact that the bars extend in opposite directions makes comparisons difficult.

By: Paula

Paula — Wed, 22 Jul 2009 01:30:48 +0000

In addition, I have seen a tornado chart tutorial by you, but I cant get mine to look like sarah’s with the split axis with 2 zeros. thanx.

By: Paula

Paula — Wed, 22 Jul 2009 01:17:42 +0000

I might be dense, but I cant find the tutorial that Sarah would have used to create the the tornado graphs with the x-axis down the middle. Can you point me to the link or are these in books. thanx

By: Jon Peltier

Jon Peltier — Fri, 16 May 2008 12:17:30 +0000

Sarah – Those are interesting graphics. Breaking down the bars into prportions (i.e., by education level) complicates the issue if you want an overlaid comparison. Sometimes I find it better to use several charts. For example, I might do a master dot/line plot of total population M/F. then a version of this for each education level, then maybe a line/dot plot for each gender showing all education levels, unstacked. Then I’d pick the few charts that supported the story I was telling, and place them all in an appendix.

By: Sarah

Sarah — Thu, 15 May 2008 21:23:13 +0000

I’m a couple of months late on this, but I like the idea of folding a population pyramid onto itself to more easily compare male to female. I have a slightly more complex case (which, by the way, I created using your tornado chart tutorial):

or if that html doesn’t work: http://epdc.org/CPGraphsLARGE/HumanCapitalPyramidNEW/Bolivia_cur.png

I guess one option would be to turn all of the series into lines, but since we have age ranges and not single-year ages, I think I prefer bars. But it’s difficult to compare male to female education levels except in the most pronounced cases, such as Afghanistan:

http://epdc.org/CPGraphsLARGE/HumanCapitalPyramidNEW/Afghanistan_cur.png

By: Kevin Johnson

Kevin Johnson — Sat, 15 Mar 2008 06:05:44 +0000

Thanks Jon.

By: Jon Peltier

Jon Peltier — Sat, 15 Mar 2008 02:28:53 +0000

Hi Kevin – There’s not necessarily any significance between the left andd right sides of the tornado plot. You have two opposite items being compared so you draw one in each direction. Often tornado charts are used to compare effectss of separate categories, rather than of an ordered set of categries (like the ages in my example). In the case of independent categories it’s customary to rank them from the widest spread at the top to the narrowest at the bottom. But you don’t sort each side, you sort the categories by the total of both measurements in each. I’ve also used charts which weren’t charts showing two populations, but more like floating bars going from a minimum to a maximum for each category. We sorted those from widest to narrowest and called them tornados too.

By: Kevin Johnson

Kevin Johnson — Fri, 14 Mar 2008 23:31:18 +0000

Can someone tell me what the right and left side of a tornado chart means. I am having problems understanding it. For instance, my text book says that the cell that has the most impact on the EMV should be placed at the top of the chart. However, my problem involves setting several input cell to + and – 20% of there base case. Nevertheless, two of my imput cells have an amount of -40,000 but yet both are shown at different locations on the chart. Could someone explain. Thanks.

By: Jon Peltier

Jon Peltier — Mon, 03 Mar 2008 22:48:13 +0000

Doug – Hat’s off to Jorge, I think I prefer the middle one. You’re right about the line chart: it isn’t really a time-line, but it causes the reader to think it shows a continuous span of years.

By: Doug Glancy

Doug Glancy — Mon, 03 Mar 2008 22:41:23 +0000

Jon, I like the first one. I find it captures the chronology well, because it reads left to right, it’s easy to compare each pair, and the new labels are very clean. The dot plot doesn’t quite jibe for me, because when I see one I expect it to be over years, e.g., from year 1980 to 2005. Thanks.