
can be added to obtain a minimum design for a quadratic model, which has a
condition number of 6.3 (see Figure 3.9).
The 2-D full factorial design with three levels shown in Table 3.5 has a
better condition number (4.4) for a quadratic model.
3-D designs
In three dimensions, the four point design of Table 3.6 is best for a linear
model.
It can be expanded to a full factorial design (see Table 3.7), which is
appropriate for an interaction model.
Also at three dimensions, the star design shown in Table 3.8 is less
accurate and requires more work than the minimum design of Table 3.6.
However, combining the centred star design with the full factorial design
of Table 3.7 gives a so-called composite centred design, suitable for a quad-
ratic model, having a condition number of 4.4. If fewer points are wanted,
Figure 3.9 Minimum design for a 2- D quadratic model
Table 3.5 2-D full factorial design with three leve ls
No x y No x y
1 1 16 10
201711
3118 01
4 109 11
500
Table 3.6 Minimum 3-D design for assessing the coefficients of a linear model
No x y z
1 1 11
211 1
3 111
4111
Age of Air and Ventilation Efficiency 55
Kommentare zu diesen Handbüchern