Morris method
The
guiding philosophy of Morris method is to determine which factors may
be considered to have effects, which are negligible, linear and
additive, or non-linear or involved in interactions with other
parameters, see reference [13].
The experimental plan is composed of individually randomized
‘one-factor-at-a-time’ experiments, in which the
impact of changing the value of each of the chosen factor is evaluated
in turn.
The number of model executions is computed as r = (k+1), where r is the number of
trajectories (successions of points starting from a random base vector
in which two consecutive elements differ only for one component) and k the number of
model input factors.
For each factor, the Morris method operates on selected levels. These
levels correspond to the quantiles of the factor distribution. In
particular:
For 4 levels, the 12.50th, 37.50th,
62.50th and 87.50th quantiles are taken.
For 6 levels, the 8.33th, 25.00th, 41.66th, 58.33th, 75.00th, 91.66th
quantiles are taken.
For 8 levels, the 6.25th, 18.75th, 31.25th, 43.75th, 56.25th, 68.75th,
81.25th and 93.75th quantiles are taken.
Morris estimates the main effect of a factor by computing a number r of local measures, at different points x1, x2, .., xr in the input space, and taking their average (this reduces the dependence on the specific point that a local experiment has). These r values are selected such that each factor is varied over its interval of experimentation. Morris wishes to determine which factors have negligible effects, linear and additive effects, non-linear or interaction effects, see reference [13]. The Morris method doesn't allow the UA analysis.
IMPORTANT: The input factor list has to contain uncorrelated items