Read through the assignment before starting!
A study was performed
to determine the amount of heat evolved during the hardening
of certain cements, as it depends on the percentage of 
four constituents. 

C3A C3S CAF C2S HEAT

7.0 26.0 6.0 60.0 78.5
1.0 29.0 15.0 52.0 74.3
11.0 56.0 8.0 20.0 104.3
11.0 31.0 8.0 47.0 87.6
7.0 52.0 6.0 33.0 95.9
11.0 55.0 9.0 22.0 109.2
3.0 71.0 17.0 6.0 102.7
1.0 31.0 22.0 44.0 72.5
2.0 54.0 18.0 22.0 93.1
21.0 47.0 4.0 26.0 115.9
1.0 40.0 23.0 34.0 83.8
11.0 66.0 9.0 12.0 113.3
10.0 68.0 8.0 12.0 109.4

Compare the 16 possible models corresponding to 
include/not include for each of the four independent
variables (HEAT is the dependent variable) with
respect to R^2, R^2_a, C_p, and PRESS.
Do your answers agree across methods?
You should produce plots with p (ranging from 0 to 4)
on the x-axis and your appropriate criterion on the
y-axis. You may want to truncate the plot in the case
of C_p to focus on those closest to the ideal line. 
With regard to C_p, is there one factor whose presence
or absence seems to make a large difference? Is this
something that we might have been able to spot with
a stepwise procedure?

Hint: You may want to read through the part of Venables and
Ripley devoted to model selection before implementing
this. You should also note that Splus uses a different
definition of C_p than the one we've been working with.