Supplementary Exercise 2.69 of IPS7e
------------------------------------

Positive association found among 22 companies that announced large layoffs in 1994, 
between the number size of the layoffs and the compensation of the CEOs.
(By the way, to call 0.31 a strong correlation is very unusual, it would normally
be termed at most a moderate correlation.)

This association could be explained as a common response to the size of the 
company (measured by its number of employees). This means that the size affects
both variables of interest (layoff size and compensation). Therefore the company 
size is a confounding variable and there might be no real association between
layoff size and compensation.

Let
  y = compensation of the CEOs
  x = size of layoff
  z = company size

We could then draw a diagram with the arrows:
- arrow from z to x (presumably positive association),
- arrow from z to y (positive association),
- but no arrow between y and x.

It is also possible that a relation exists between y and x, but one would 
need to "control for" (epidemiology terminology for "account for") the confounding 
effect of z in order to get a correct estimate of its strength and direction.