[Phys-L] tack-tossing

[John Denker <jsd@av8n.com>]

On 06/08/2015 09:57 PM, I wrote:

> I have been asked on various occasions how I would teach people about  
> climate change.  I do *not* recommend attacking the issue head-on.
> Instead I suggest starting with tack-tossing
>   https://www.av8n.com/physics/tack-tossing.htm
> or some similar ultra-simple probability exercise.  The point
> is, people need to have some clue what real data is /supposed/ 
> to look like.  The tack data is noisy;  it takes a looong time
> for the running average to settle down to any kind of asymptote.

I recently added a new section to my tack-tossing writeup.
It uses tack-tossing data to construct a random walk.
  https://www.av8n.com/physics/tack-tossing.htm#sec-random-walk

This is useful as example of data with a long-term trend 
plus fluctuations.  There are a lot of people in this world
who don't have anywhere near enough experience looking at
noisy data.

 ++ If you look at a long-enough run of data, the trends
  become clear.

 -- Beware of selecting the data. When the data is noisy, 
  swindlers can always pick out a subset in such a way as 
  to “prove” anything they like. To defend against this, 
  demand to see all the data. If what they say about part 
  of the data is not true of the dataset as a whole, you
  know you’re being swindled.


Pedagogical considerations:

 *) Tack-tossing is a super-easy experiment. It is a cheap 
  way of generating tons of data.
 *) Unlike a coin, we don’t expect the tack to have equal
  probabilities for heads and tails.
 *) The tack-tossing data is real data unto itself, but 
  it also serves as a metaphor for other processes that 
  exhibit both trends and fluctuations.
 *) If necessary, you can supplement the real tack-tossing
  data with computer simulations. This is a super-cheap way 
  of generating additional tons of data. However, you should
  start by gathering enough 100% real hands-on data so that
  everybody can see that the simulation behaves the same as 
  real tacks.