Supplementary Exercise 9.40 of IPS7e
------------------------------------

Many solutions are possible, and they are all valid (provided the table 
indeed shows no association).

The simplest, but not too interesting examples are when all counts of
the table are identical. Only slightly more interesting are situations
where the three rows (or columns) are identical. In such cases, it is
obvious that the marginal and conditional distributions are the same.

Further complicated again are tables where the numbers of the second row are 
twice those of the first row, and the third row twice those of the second
row. For example,

5  10 25
10 20 50
20 40 100

Also here it is intuitively obvious that the marginal and conditional
distributions are the same. One could also compute the expected values for
each cell and check that they are equal to the observed values. For example
for cell (1,1):
e11 = 40*35/280 = 5, where
- the sum of the first row = 40
- the sum of the first column = 35
- the total sum = 280.

Any table where all the rows are multiples of the first row (say), has no
association between row and column variables.