Beware of charts that are trying to make you outraged, afraid, or sell you something

We’ve all seen them:  Articles with charts expressing how “We really need to do something about A!”  Any moron looking at this chart can tell you that A causes B.  Case closed.  Double burned.  Owned.  Fact-checked.

To use an example from the link:  We really need to eliminate U.S. spending on science, space, and technology!  Any moron can tell you that U.S. spending on science, space, and technology causes suicides by hanging, strangling, and suffocation.

Or maybe it’s suicides by hanging, strangling, and suffocation causing increased U.S. spending on science, space, and technology?

For a decade, these were correlated over 99.7%.  That correlation level is unheard of – even for stuff we know for certain is correlated 100%.

Correlation does not equal causality.  To prove causality, you absolutely need correlation.  This is true.  However, just because you have correlation doesn’t mean anything at all.  You have correlation eh?  Good for you.  You want a cookie with that correlation?

To believe correlation = causality is to believe everyone with the opportunity to murder someone must have murdered them.  They all did it:  The coworker,  The FedEx driver.  The spouse.  The gas station clerk.  Colonel Mustard.  All of ’em.

The media loves to play this trick on people.  Just because someone is on TV, wears nice clothes, speaks well, is passionate, and may even be attractive, doesn’t mean they understand statistics – at all.  Every reporter should have to learn this, but I’m afraid almost none do.

It may be too harsh saying they are “playing tricks,” because many truly believe what they are reporting.  Many are even passionate about it.  When people believe the lie they are propagating, that makes it all the more convincing for the rest of us.

You don’t have to be a good actor when you actually believe the lie you are reporting.

Most charts are trying to make you angry, afraid, or sell you something.  This correlation and causality can fool you – especially when they seem somehow related.

I read this somewhere but can’t remember who to credit, but here is the paraphrase:

Lets say you see a shocking graph showing a high correlation of people who live near high power transmission lines with cancer.  Therefore:  Power lines cause cancer.  Sorry – not so simple:

High power transmission lines are considered an eyesore.  This is not controversial.  Therefore, houses near high power transmission lines sell for less that those without them.  Lower income people generally buy lower cost houses.  This is also not controversial.  Lower income people, on average, also have more unhealthy habits that higher income people, on average.  I’m not saying all, person about to “burn” me with the all caps quadruple exclamation point comment.  Take it easy:  Step away from the caps lock button.

So is it the unhealthy habits or the power lines (or something completely unrelated to either) that’s giving them cancer?

You need more data to know anything.  You need to look at multiple variables to know anything.  A single variable “study” is for suckers.  Your one, single variable chart doesn’t prove a darn thing.

If your variable is people, then it gets even tougher.  Each person has a unique history.  Say you are measuring the impact of (for example) drug A on health characteristic B.  Group C is on placebo and group D is on drug A.

Group C has many people with unhealthy habits.  Group D has few people with unhealthy habits.  The people running the studies ask about health habits though questionnaires, but can’t possibly cover them all.  The people answering the questionnaires may be honest or they may lie.  The people answering may not remember.  Each person has a unique, little known  genetic history and lives in unique, little known environmental conditions as well.

In the end, the group on drug A (group D) had 3x as many favorable results as group C, on placebo.  Therefore:  Drug A is 300% as effective as placebo at improving health characteristic B.  “Ask your doctor about expensive drug A!”


Comparing one person to the next to try to predict what effect A has on their health, behavior, or other trait is challenging to put it mildly.  If a human study is not also double blind, then it is utterly useless.

I’m not saying all medical studies are useless, but you need to look at far more than a single variable to determine anything at all.  Show me a dozen different variables.  They better all point the same direction.  The variables studied on one should ideally have financial interests opposite the financial interests of the variables studied on the other.

When that happens, and they’re both pointing the same direction, then you’ve got my interest.  Then I might consider investigating further.  Until you show me this, I don’t care what your study says.

Also – You must be able to document, for each data point:

  • Where was it gathered?
  • Was the instrumentation collecting each data point the same as the rest?
  • Was the instrumentation calibrated?
  • Did the environment around the instrumentation change over time?
  • If the environment around the instrumentation changed over time, does it impact the data?
  • What is the error band around each data point?
  • Are you are looking at raw data or smoothed data?
  • If smoothed data, how was smoothing applied?
  • Were there any “adjustments” performed?
  • If so, why was the data “adjusted?”

And in general:

  • Who paid for the data to be collected?
  • Does the data collector have a vested interest in the data pointing a certain direction?
  • Is the data collector employed by someone who has a vested interest in the data pointing a certain direction?
  • Who benefits (or is harmed) if I believe this?
  • Who benefits (or is harmed) if I don’t?
  • Why was the starting date chosen (for data changes over time)?
  • Why was the end date chosen (for data changes over time)?

Until you know all these things, you really don’t know anything about that chart.  You have absolutely no business being outraged by it, let alone protesting about it.

You don’t know anything and neither do I.  Quit being a sucker who believes you actually do.

Do you know who’s smarter than you?  You – 5 years from today.  Do you know who gets it wrong more often than not?  Most people.  Even most “smart” people.

See:  Salem witch trials.  Bloodletting.  Leeches.  Cocaine as a miracle drug.  The Nazi party.  All manner of religious persecutions over time.

The cool kids are stupid.  Stay away from them.  Don’t even let the smart kids tell you what to think.  They get it wrong an awful lot.  Think for yourself.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s