Column Chart Extends Beyond The Scale [Chart Review]
[source]
Hypothetically, If you were presenting this data:
- Would you have used a line graph instead of the column chart?
- Would you have started the y-axis at a value greater than zero or tried to break the scale to magnify the range of 35-45?
In the print edition of BusinessWeek where this chart appeared, the title read, “The number of people in the U.S. without health insurance has risen by nearly 18% – faster than the overall population”. It would have been interesting to see the change in the population and this chart plotted together in a line graph. Or, if you are going to make a statement like the one above, at least note the percent change in population as a reference point.
Another interesting design choice is the fact that the data in 2006 extends beyond the max value of the y-axis. The scale should have been set at a maximum of 50 for this chart so all of the columns were within the range.
December 26th, 2008 at 9:33 am
If the purpose is to compare uninsured with total population, there are a couple options that make more sense than a column chart showing only uninsured.
1. Line chart with two series: annual population divided by 1999 population (i.e., 1999 = 100%) and annual uninsured divided by 1999 uninsured. This approach is very sensitive to starting year (note that uninsured went down between 1999 and 2000). Data labels on points (or on min and max points) would add information.
2. Line or column chart with one series: % uninsured (uninsured population divided by total population).
If I were stuck with a chart showing only uninsured population, I’d probably use a line chart, and for these numbers I’d probably use a Y axis scale of zero to 50M
December 26th, 2008 at 10:45 am
Jon – I agree that with a little more data, like total population, there could be a more informative chart rather than just total uninsured from ’99 – ’07.
Why would you not just magnify $35M to $50M with an axis break so the changes are more visible?
Thanks for the comment! I hope you had a nice Christmas.
December 26th, 2008 at 1:16 pm
No, I don’t think every graph always has to have at least one gridline beyond all the data. I think it’s okay to have data that extends a tiny bit past the last grid line.
For a scale without gridlines, it’s a little less forgivable, because it’s harder to gauge the last point by eye that way. And here, there was no reason to avoid the fifty, which would have ben a fine round number to use (and which could then have allowed an interval of ten per label, instead of the excessive five per label.
December 26th, 2008 at 1:37 pm
Tony -
If the changes are relatively smooth, as in this case, I don’t think it’s worth gaining the added resolution of a 35-to-50 scale by removing the clear indication of the entire 0-to-whatever value of each point. Yeah, I know, with markers (as opposed to bars) you don’t need to start at zero.
Remember, showing just this data, unnormalized, was not one of my preferences in the first place.
December 26th, 2008 at 2:41 pm
@Derek – In this case, I think it makes perfect sense to go up to 50 on the y-axis with ticks in increments of 10. Good suggestion.
@Jon – If I was only allotted the current data, I probably would have gone with a line chart. I think most people look at line charts in terms of time scale on the x-axis and bar charts as comparing “things”. I am generalizing and just drawing my own conclusion based I what I initially think when I look at these types of charts.
Another option is to use a line chart with my single data point example showing the overall percent change from ’99 to ’07. This way you can see the changes each year and also have the total change all in one line chart.
January 1st, 2009 at 6:35 am
I do agree with derek.
Sometimes, I want to let the column of the big numbers exceed the bigest gridline,but I dont know how to make out this in excel ,anyone can tell me ?
January 2nd, 2009 at 8:36 am
Liu -
In Tony’s example, if you set a minimum of 0, a maximum of 48, and a spacing of 5, the axis will go up to 48, but the last tick mark and gridline will be at 45.
January 2nd, 2009 at 9:01 am
Jon －
Thank you.
by your method ,there is still some difference with the effect I want
January 2nd, 2009 at 1:31 pm
Liu -
Supply a few more details, perhaps also a link to a mock up, and I’ll have a look.
January 2nd, 2009 at 9:41 pm
Jon:
thank you for your reply
I want the effect like this:
http://junkcharts.typepad.com/.a/6a00d8341e992c53ef010535e493e9970b-320wi
I can make it out by cell,text box, but is there any better method ?
January 2nd, 2009 at 10:27 pm
That’s more than skipping the last gridline and the last axis tick. That’s having one bar that’s twice as tall as the rest of the chart.
To get this effect in Excel, you would make the chart tall enough for all of the data, then use a few dummy series to draw the box outlining the nominal plot area and the portion of the Y axis that is shown.