Sunday, February 14, 2016

Correlation ≠ cause and effect

I was recently directed to a lovely example of how an observation of correlation can be misinterpreted with regard to cause and effect.*  It comes from Mind of the Raven by Bernd Heinrich.  Here's the excerpt:

At dusk on September 7, 1997, a cougar crept up on Ginny Hannum as she was working at the back of her cabin at the head of Boulder Canyon in Colorado.  The cougar crouched low among the rocks, facing her from about twenty feet, and it was ready to pounce.  

Although Mrs. Hannum was unaware of the cougar's presence, she had become "somewhat annoyed" by a raven "putting on a fuss like crazy.  The noisy raven kept coming closer, having started its commotion twenty minutes earlier from about three hundred yards away.  Was this raven trying to say something?  She started to listen more closely.

The cougar was ready to make its kill, but the raven was close, and it made pass over the woman, calling raucously, then flying up above her to some rocks, where she finally saw the crouching cougar.  As the cougar glared down with yellow eyes locked onto hers, Hannum quickly backed off and called her three-hundred-pound husband.  The surprise attack had been averted. She had been saved.  "That raven saved my life." The event was declared a miracle in the news.

A miracle is any event the natural cause of which we do not understand.  Why did the raven call?  To the religious Hannums, it seemed a miracle that a raven would go out of its way to deliberately save a human life.  To me, raven behavior is still a miracle, although I have faith that this raven's behavior was within the realm of what ravens normally do.  They are alert to predators that could potentially provide them with food.  Perhaps the raven had been luring the ion to make a kill, alerting it to a suitable target.  If the lion had feasted, so would the raven.  That is, both would have benefited, as expected in communication.

After presenting some more evidence, the author concludes:

Everything I know about ravens, as well as folklore, is congruent with the idea that ravens communicate not only with each other, but also with hunters, to get in on their spoils.

Whatever else these to incidents illustrate, they show the difficulty of interpreting communication and how much communication can depend on the mind-set of the receiver.  To make sense of communication, the first relevant questions to ask are: What are the costs and the payoff to the givers and potential receivers of the signals given.

All in all, this is an excellent example that shows the dangers of making quick assumptions from a few observations without applying rigorous thinking to possible underlying causal factors.

--

Many thanks to Dr. Rod Phillips.

3 comments:

beverly said...

Excellent advice, particularly during an election year. I would only add that also, we receive the communication that we want to receive. I recently had a disagreement with someone on Facebook regarding whether Janet Yellen "liked" the idea of negative interest rates, based on our starkly differing interpretations of her recent congressional testimony.

Dominic DeMartini said...

Really great article, Paul. Can't tell you how many times in Biostatistics and Epidemiology last year that we heard "correlation is not causation!". Cool to see it somewhere outside of class!

Michael said...

Nice story Paul. Parallels in health between serum cholesterol and atheroma. Cholesterol is almost always involved in atheroma just as firefighters are almost always at a serious fire. The correlation was labelled causative many years ago and this simplistic conclusion still permeates our policies and behaviours, delaying the move to lower carbohydrate diets.
On another front... there's a recent book called "This idea must die" and the idea of "cause and effect" is one that the author identifies as delaying our understanding of some issues. I haven't read it, but it's a tantalising thought. I remember being stunned in a narrative therapy training session when I was called on the limitations of my causal thinking in my approaches to counselling!