Sunday, December 9, 2012

What they conceal is vital - II


In this list of Carl Sagan quotes, I saw the following quote:
"I'm often asked the question, "Do you think there is extraterrestrial intelligence?" I give the standard arguments -- there are a lot of places out there, and use the word *billions*, and so on. And then I say it would be astonishing to me if there weren't extraterrestrial intelligence, but of course there is as yet no compelling evidence for it. And then I'm asked, "Yeah, but what do you really think?" I say, "I just told you what I really think." "Yeah, but what's your gut feeling?" But I try not to think with my gut. Really, it's okay to reserve judgment until the evidence is in.
But people often rely on gut feel for making decisions. When news reports have a lot of numbers, eyes typically glaze over so people in power get away with saying anything.What Steven Pinker calls 'failure of statistical thinking' has made it difficult for Facts to survive. In this video, Daniel Kahneman distinguishes two types of thinking: System 1 and System 2.System 1 represents what we call intuition. It tirelessly provides us with quick impressions, intentions and feelings. System 2, on the other hand, represents reason, self-control and considered decision making.

System  1 is fast and does not require much effort. System 2 is slow and requires effort. We rely most of the time on System 1 for our regular activities and it does fine.Occasionally, this causes problems. There are times when some statistical thinking using System 2 would have been beneficial but we often skip it since it requires time and effort. Advertising, political, nationalistic and religious messages target System 1 which is why they are so effective.

Some years back, I had read a book called How to Lie with Statistics by  Darrell Huff, the most widely read statistics book in the history of the world.  You can read it online here, Written almost 60 years ago, it gives many ways in which people present statistics in order to favour their biases. He says that the book sounds like a how-to manual for crooks for which his justification is - "the crooks already know these tricks; honest men must learn them in self-defence." Some of the techniques that he discusses are:
  1. Using a sample with built-in bias. Examples are TV and Internet polls.
  2. Using any of mean, median or mode as the 'average' depending on which one best represents your bias. Since most people won't know the difference between them ,you will generally be safe. For example, to depict the average income of the inhabitants of a country, the median is more informative than the mean. There is a joke that when Bill Gates visits an old age home, all inhabitants are millionaires on average.
  3. Failing to mention some numbers like sample size, confidence intervals etc. This is especially a problem on TV  because the screen changes so fast that you don't have time to read everything and only the shape of the image stays  in your mind.
  4. Misleading graphs - An example
  5. Misleading  figures - An example.
  6. Comparing  wrong percentages - An example
  7. Correlation is not causation - An example
  8. Post hoc ergo propter hoc
  9. Using impressively precise figures - Saying that the monthly expenditure of an average family is Rs. 12436 sounds more authoritative than saying it is around Rs. 12,000.
  10. Beware of extrapolations.
Darrell Huff says that there are 5 questions that one must ask when one sees any statistic:
  1. Who Says So?
  2. How Does He Know?
  3. What's Missing?
  4. Did Somebody Change the Subject?
  5. Does It Make Sense? 
Whenever one political party levels an accusation against a member of  another party, the latter responds by saying something like, 'People will not believe such wild allegations. They know better.' Want to bet? Initial reports tend to persist and myths are remarkably easy to spread because we use System  1 much more than System 2. This makes us us more enamoured of superficial things as depicted in Chekhov's short story, The First-Class Passenger. This is a difficult trap to avoid.

PS: There are many examples of statical machinations dissected at Ben  Goldacre's blog.

No comments:

Post a Comment