Data-driven fallacy | Hacker News Reader

Data-driven fallacy

cutchai.wordpress.com

48 pointsmycodebreaks18 comments

Manipulating the scales on graphs is one of the oldest tricks in the book. It's one of the many deceitful practices that Edward Tufte talks about in The Visual Display of Quantitative Information. In fact, in the chapter on "Graphical Integrity", he writes: "For many people the first word that comes to mind when they think about statistical charts is 'lie'."

If I were interested in investing in Twitter stock, I wouldn't be nearly as concerned with the growth of their user population as I'd be in the growth of their bottom line (profits). Oh, sorry, there are no profits.[1] As they say, "you can't make up losses on volume".

[1] https://en.wikipedia.org/wiki/Twitter#IPO

I know very little about economy, but "you can't make up losses on volume" does not sound right. Let's say you have fixed costs F per year and you make a profit of P per user. Then as soon as numUsers * P >= F, you made up your losses on volume.

> you make a profit of P per user

I believe the point of the quote is that they're making a loss per user

These are not losses. Losses are, as in Twitter's case, when P is negative.

GP is surmising the case where marginal profit per user is positive, even though average profit per user is still negative.

IOW, calling it I, net income per user, would have made the argument more clear.

By golly you are right! They just need a complex count of users and negative complex profit and the result is positive!

How to Lie With Statistics (1954) predates Tufte by quite a bit.

If you really need to clip off the bottom of your bar graph for clarity, jagged lines do a wonderful job of explaining that instantly. I don't see the point of a bar graph that doesn't start at 0 or indicate an omitted range, except to deceive.

That's being too generous. NEVER clip off the bottom of your bar graph. Use something that isn't anchored to the axis - dots instead of bars which then put the focus on relative position. A bar draws the reader to comparing the relative length of the bar, so if "twice as long" doesn't mean "twice as much", it's terribly deceptive.

I see your point. Though how does your advice scale to logarithmic scales? (I.e. what's a general reason to allow logarithmic scales, but not affine ones?)

If you're using a logarithmic scale you shouldn't be using a bar chart, for the exact same reason as above - when you see bars, you naturally compare them on length. If the scale is logarithmic, then the form of the graph is in conflict with the data, and the graph becomes confusing. You should strive for every part of the graph to be in harmony and to repeat the same message.

The following paper on lucidity in scientific communication has been a big influence on my thoughts. Briefly, the author stresses that you should aim for lucid pattern repetition over gratuitious variation, i.e. the repetition of the same idea communicated in similar ways, over variety for the sake of appearing to be superficially interesting.

http://www.atm.damtp.cam.ac.uk/mcintyre/papers/LHCE/lucidity...

Thanks!

As a brief addendum... The point of a log scale is to be able to visually compare the relative percent rates of change - last quarter we had 50% growth. This quarter it's less... Dots and lines are useful for this... Not bars.

Really?

How about if you're trying to show in detail a small change in a large number? If the scale started at zero, all bars would look equal.

Don't use a bar graph, then.

One option is to plot the change itself.

Notice also that they plotted exactly 6 bars.

Choosing the start point is also a way to magnify perceived growth-- perhaps the growth was much stronger before that and is actually levelling off? Of course, if they had drawn 8 bars I could have said the same, so I'm not accusing them of anything; I'm simply pointing out the possibility.

What a pretentious title to something newspapers do all the time and everyone learns about in school when talking about source criticism.

Reply on news.ycombinator.com