Supplementary Exercise 12.25 of IPS7e
-------------------------------------

A total of N=10+5+11+10=36 observations, and 4 groups.

Model: X_ij are normally distributed N(mu_i,sigma) and independent.

(a) 
Source of	Degrees of	Sum of		Mean	  F	  P	
variation	freedom		squares		square
-----------------------------------------------------------------------
country		3		104,855.87	34952	  15.86	  <0.001
error		32		 70,500.59	 2203
-------------------------------------------------------
total		35		175,356.46

For calculation of the P-value, see (c) below.

(b) H0: all (population) means are equal, and Ha: not H0, or some (population) means differ

(c) Table F in PSLS (or Table E in IPS) gives the 99.9%-percentile of
F(3,25), our best conservative approximation to F(3,32), as 7.45, corresponding
to P<0.001. Using statistical software, we can determine the exact value in F(3,32) 
as P=1.7*10^-6, or 0.0000017. There is no need for all those decimals, so P<0.001 is
perfectly fine as the stated P-value.

We conclude there is strong evidence against the null hypothesis, or conversely
evidence in favour of some means being different. 

(d) The pooled variance s_p^2 can always be read off the table as the MSE:

s_p^2=2203, s_p =sqrt(2203) = 46.9