Misleading charts and data

[Richard Schollar]

%s are always a good target - you only get sensible data when a consistent baseline is used. Consider:

Politician: GDP fell last year by 20% but this year we have seen a rise of 25% so we are better off than ever before!

Year__GDP__GDP movement
Year0_100________-
Year1_80_______-20%
Year2_100______+25%

Net improvement of year2 compared to year0: 0%

 

[Colin Legg]

%s are always a good target - you only get sensible data when a consistent baseline is used. Consider:

Politician: GDP fell last year by 20% but this year we have seen a rise of 25% so we are better off than ever before!

Year__GDP__GDP movement
Year0_100________-
Year1_80_______-20%
Year2_100______+25%

Net improvement of year2 compared to year0: 0%

It's that darn volatility drag again!

 

[PaddyD]

"%s are always a good target - you only get sensible data when a consistent baseline is use"

Exactly the point re percents of percents. Consider the following:

Widget   Y1         Y2       % growth
A           5           50          900%
B         10,000    10,001     0.01%
C          10           10           0%
D          100        200          100%

..the comment 'on average our sales grew 250% last year' is in some sense numerically correct but, from a business perspective, meaningless at best.

 

[Long Nose]

I find this confusing, although I can explain it.

<a href="http://s579.photobucket.com/albums/ss234/detwd0hb/?action=view&current=Chart.jpg" target="_blank"><img src="http://i579.photobucket.com/albums/ss234/detwd0hb/Chart.jpg" border="0" alt="Example"></a>

<a href="http://s579.photobucket.com/albums/ss234/detwd0hb/?action=view&current=ChartII.jpg" target="_blank"><img src="http://i579.photobucket.com/albums/ss234/detwd0hb/ChartII.jpg" border="0" alt="Example II"></a>

 

[SydneyGeek]

Ah... the "Unwrapped Toffee" chart. Always easy to interpret!

I agree about pivot tables. Quick summaries, easy to filter data, not so easy to lie about it.

I regularly include a checksum row at the bottom of important summary sheets, so you can see if anyone has hidden rows of unwanted numbers.

Charley Kidd has some useful hints on business charting at http://www.exceluser.com
He's also a fan of Tufte.

Denis

 

[Jon von der Heyden]

Long Nose
That is a horrible chart. I wouln't event know where to begin.

Denis,
Thank you for the pointer to exceluser.com. And thanks for the checksum tip, I think that's quite an important tip.

Paddy and Richard
Thank you for the good examples. I see what you mean.

 

[gardnertoo]

The king of the "Misleading Chart" kingdom must be the radar plot. I have used this chart for exactly one good application: graphing temperatures reported by sensors which are physically in a circular arrangement.

In this example, you are looking upstream at the thermocouples which monitor exhaust temperatures from a combustion turbine. Sensor #4 saw the most dramatic cooldown resulting from a flame-out in one combustor. The bold red series at timestamp 7:03:50 marks the time the unit was shut down by it's protective systems. This display format gives the maintenance and repair crew a good idea where to go look on the physical machine.

 

[SydneyGeek]

You're missing the point Gardnertoo, that's actually informative!

Something else that I hate, and never use, is pie charts. They often tell you nothing and can easily get overcrowded.

Denis

 

[hatman]

1) Second Y-Axis: very effective in many engineering applications when plotting two curves in drastically different units or scales on the same chart. Here's a typical example of one of mine: Time along the X-Axis, Pump RPM's from 0 to 5000 on the primary Y-Axis, and Pump Delta-Pressure from 0 to 10 psid on the secondary Y-Axis.

2) My tip for making data easy to interpret is to chart it clearly. Seriously: raw numbers are almost impossible for anyone who is not personally collecting or massaging the data directly to be able to interpret. (okay, so that sentence was even more obfuscating than many of the unclear chart examples... sorry) What I mean to say is that the person who works directly with the data sets may be able to see meaning in the raw numbers, but someone like an upper level manager who is trying to get an exectutive summary is only going to infer meaning if it's somehow presented graphically. That being said, I do have a few pieces of advice for making it easier to see trends when looking at raw numbers. The first is to specify the number of decimal points and/or or sig figs discreetly and uniformly for all cells in the column. The second is to eliminate as many insignificant figures as possible. Third, when eliminating insignificant figures, choose wisely whather you allow Excel to do it's default rounding, or implementing some other method, like Int(), Ceiling() or Floor()... all of these mathods have pros and cons, but depending upon your data you can often clarify or confuse by choosing one over the other (and sometimes, it just doesn;t matter).

3) 87.3% of all statistics are made up on the spot.

4) "There are three kinds of lies: lies, damned lies, and statistics."

5) Know when to use a log or a log-log scale... you can really skew your data by plotting it on a linear axis when it belongs on a log-scale, and vice versa.

 

[Jon von der Heyden]

3) 87.3% of all statistics are made up on the spot.

I'm definitely using this!
Seriously, thank you for your contribution.

Something else that I hate, and never use, is pie charts.

Cannot agree more! Thanks Denis