Statistical modelling

Doing the same steps as for fitting the quadratic function we obtain:

Source of variation	d.f	s.s	m.s.	v.r.
DAMAGE7	6	445.076	74.179
DAMAGE4	3	378.859	126.286
Remainder	3	66.217	22.072	4.49
Residual	683	3357.495	4.916

Here the remainder mean square is significantly greater than the residual mean square (F_3,683 = 4.49; P<0.01), so, the reduced number of categories is not as good a representation of the association with age as that with seven categories.

We decide to use the quadratic relationship with DL and DQ in our final analysis.