LISTING 10 describes the computation of the discrete model associated to a given design, and provides an estimate of its VRF. The formula used is not the best possible (the expectation of a quotient is not the quotient of expectations), but in any case expresses the fact that our ability to predict the future is necessarily less than our ability to boldly predict the past (http://en.wikipedia.org/wiki/Split_infinitive).
LISTING 11 allows the user to view the "discrepancies when predicting the past". Plotting these discrepancies against the values obtained, as chosen in this routine, allows to determine the zone of better validity of the model. Plotting these discrepancies against the experiment number can be useful to detect a shift of the process.
Finally, LISTING 12 allows us to visualize the influence of the factors. Given that the aim is to compare amplitude, these influences are not translated as it should be when the design is not balanced.
![\begin{algorithm}
% latex2html id marker 1613
[H]
\begin{algorithmic}[1]
\par
\F...
...nd{algorithmic}
\par
\caption{Discrete model and its VRF
}
\par
\end{algorithm}](img373.png)
![\begin{algorithm}
% latex2html id marker 1626
[H]
\begin{algorithmic}[1]
\par
\D...
...
\end{algorithmic}
\par
\caption{Remaining discrepancies
}
\par
\end{algorithm}](img374.png)
![\begin{algorithm}
% latex2html id marker 1634
[H]
\begin{algorithmic}[1]
\par
\D...
...
\par
\end{algorithmic}
\par
\caption{Factor influences
}
\par
\end{algorithm}](img375.png)