Junk Charts

Monday, January 23, 2006
Kaiser Fung has declared war on chartjunk.

Lots of good stuff in the archives.

11 comments:

RogerA said...

One of the Dept of Health Epidemiologists in WA state has been proselytzing for more accurate ways to present data--I fear that many folks are in the Excel mode, and just take that that program offers--clearly the presentation of data is often more "innovative" than the data itself.

Knucklehead said...

Morgan,

It seems there aren't many with even a remote interest in charts or chartjunk so we might not get many Yargbian eyes poring over the charts at the site you linked to. But it sure looks to this pair of eyes that Kaiser has made an error in his replot of the data that yields the blue line in the bottom of the two charts on the right (rescale y-axis) in Dissecting two axes.

Of course I could be missing something fundamental.

Morgan said...

knucklehead:

I don't see the error, but your eyes may be better than mine. There are two y-axes, one for each series, and he's rescaled one of them because - well, to make the point that the choice of scale on the secondary axis is arbitrary, and can be used to make the two lines appear to track one another closely (or not) as desired.

My real problem with the original chart is that it is denominated in nominal dollars, so of course both exploration and the price of oil show an exponential increase.

As the original chart is constructed, it seems to imply that capital expenditure and the price of oil are closely related.

I have to admit, though, it's not clear to me what the intent of the original graph was. The title suggests that it was only supposed to show that capital investment has increased (in nominal terms) over the years (which would make the price of oil superfluous). Hardly worth the ink.

Eric Blair said...

That's Edward f*n Tufte! You have to read his book "The Visual Display of Quantitative Information". You will never look at a chart the same way again.

You have to see his presentation on the Challenger disaster and the awful graphs those guys that made the O-rings had, as opposed to what the same data would actually have told them about the failure rate of the O-rings and launch temprature.

Morgan said...

rogera:

Part of the problem is a "take what the software gives me" approach, but there is plenty of advocacy via data presentation that goes on, too. In those cases, the goal is to eliminate the context by, for example, truncating axes or transforming the data in a way that is not clearly explained.

Very much like claiming that a slowing of the rate of growth of a government program is a "cut" in that program.

Morgan said...

eric:

I think that he only likes Tufte's work (and references it often) but I suppose "Kaiser" could be a pseudonym.

That would be something, huh?

I'll look for a copy of the book. Sounds like a great read.

Knucklehead said...

As far as I understood the piece, the blue line graphs on all three charts present the very same data.

The original chart has the most gradients marked along the y-axis. From that chart it can be seen that no value for y exceeds approximately 175.

The top chart on the right has only one gradient marked on the y-axis. That is 100. The blue line in that chart is completely consistent with the one in the original chart.

In the bottom chart on the right the y-axis is rescaled with the single marked gradient being 200. If the blue graph line plots the same data as the original chart, then no portion of the blue graph line should reach the 200 value along the y-axis let alone blow past it well into 300+ territory. It looks to me that while he rescaled his y-axis he somehow failed to let the data know about it ;)

BTW, the red graph line in the bottom right chart looks like what we'd expect with the y-axis rescaled. But the blue graph line doesn't look ANY different that it did in the chart on top of it.

Morgan said...

knucklehead:

There are two y axes, only one of which was rescaled.

The blue line is always the price of oil, and is plotted against the scale on the right side, but not shown on the re-drawn graphs (which looks like Kaiser's error to me).

The bars cum red lines are capital investment, plotted against the scale on the left side. So the capital investment axis (left side) was rescaled, while the price of oil side (right side) was not.

Hence the blue line stays the same (absolutely and relative to the unchanged right-side scale), while the red line's slope is halved (because the left-side "capital expenditure" scale changed).

It's not always inappropriate to use two axes (or at least I don't think so, I've done it here ), but it can be misleading.

Knucklehead said...

OK, then, Morgan. I didn't dig deep enough into the graphs to realize the blue line was plotted against a different Y-axis than the red line.

There are, of course, all sorts of ways to play with graphs. Some of them are borderline dishonest. I'm not entirely convinced that simply choosing the type of graph to enhance the visual impact of that which one is advocating is "dishonest".

For example, two different product units might be competing for limited budget increases for MRP presentations. One might show their percentage growth in number of units sold using green for their product and red for the "other". They might pick a bar chart because it "looks" better. The "other" group might choose a pie chart showing the portion of total revenue represented by the two products. The budgeteers should be able to sift through the visuals ;)

Morgan said...

knucklehead:

Now we're really treading the same ground that we have trod with regard to opinion versus "objective" journalism. Don't remember how that came out, but there was some opinion that in some cases advocacy is expected, while in others we expect the straight scoop.

MeaninglessHotAir said...

there was some opinion that in some cases advocacy is expected, while in others we expect the straight scoop.

Whatever the ideal is, the presence of the blogs has made it clear to me that in most of the cases we're expecting the straight scoop and getting advocacy instead. Advocacy, moreover, which claims to be the straight scoop in a holier-than-thou moralistic tone. It is insufferable.

You really hadn't read Tufte's book before? I second Eric, it's a must read.