Sensitivity Indices with Dependent Inputs
Sensitivity indices with dependent correlated inputs, using a
method based on PLS regression.
- add checks on the arguments 'options' and 'output'
- improve display of the'sivip' objects:
- rename the slot 'percentage' into 'tsivip.percent'
- add the slot 'fo.isivip.percent'
- the slots '*.percent' are displayed by the 'print' function,
when the option 'all=TRUE' is set only
- reduce time execution of function rlaz
- file NAMESPACE: add import of hidden packages
- Add the option 'fast' to 'sivipm' and 'sivipboot': a new
algorithm is implemented, more adapted to big datasets.
- When the Q2 are required, the Q2ckh are also output.
- The significant components determined by simca rule,
are calculated from the Q2 of each response variable, in addition to
the Q2T. Formerly, from the Q2T only.
- It is checked that the number of components is not greater
than the number of monomials.
- Creation of the class 'sivip', container of the result
of the 'sivipm' function. Its method 'print', thanks to its option
'all', allows to limit the amount of printed components.
Its method 'getNames' displays the names of the non-null slots.
- Fix a bug in the computation of fo.isivip:
they were calculated for the first component instead of the last one.
- Fix a bug in sivipm : when no polynomials description is given,
only one variable was taken into account.
- New algorithm to generate full polynomials.
- Add an option 'all=FALSE' to the 'print' method.
By default, the monomials of a polynomial are not displayed.
Only their number.
- Complete rewrite of the sivipm-0.5 version dated 2012-07-31