Generalized Nataf Transformation¶
The Generalized Nataf transformation is an isoprobabilistic transformation which is
used under the following context: the input random vector is with marginal
cumulative density functions
and copula
. The copula is assumed to be
elliptical.
Introduction¶
Let be a deterministic vector, let
be the
limit state function of the model and let
be an event whose probability
is defined as:
(1)¶
The Generalized Nataf transformation allows one to calculate
. This mapping
is a diffeomorphism from the support of
into the standard space
, where distributions are spherical, with zero mean,
unit variance and unit correlation matrix. The type of the spherical
distribution is the type of the elliptical copula
.
The Generalized Nataf transformation presented here is a generalization
of the traditional Nataf transformation (see [nataf1962]): the reference
[lebrun2009a] shows that the Nataf transformation can be used
only if the copula of is normal. The Generalized Nataf
transformation (see [lebrun2009b]) extends the Nataf
transformation to elliptical copulas.
Let us recall some definitions.
A random vector in
has an elliptical
distribution if and only if there exists a deterministic vector
such that the characteristic function of
is a scalar function of the quadratic
form
:
with a symmetric positive definite matrix of
rank
. As
is symmetric positive, it can
be written in the form
,
where
is the diagonal matrix
with
and
.
With a specific choice of normalization for , in the case
of finite second moment, the covariance matrix of
is
and
is then its linear
correlation matrix. The matrix
is always well-defined,
even if the distribution has no finite second moment: even in this
case, we call it the correlation matrix of the distribution. We note
.
We denote by the
cumulative distribution function of the elliptical distribution
.
An elliptical copula
is the copula of an elliptical distribution
.
The generic elliptical representative of an elliptical distribution family
is the elliptical distribution whose cumulative distribution function is
.
The standard spherical representative of an elliptical distribution family
is the spherical distribution whose cumulative distribution function is
.
The family of distributions with marginal cumulative distribution functions
and any elliptical copula
is denoted by
. The cumulative distribution function of this distribution is noted
.
The random vector is supposed to be continuous and
with full rank. It is also supposed that its cumulative marginal
distribution functions
are strictly increasing (so they
are bijective) and that the matrix
of its elliptical
copula is symmetric positive definite.
Generalized Nataf transformation¶
Let in
be a continuous random vector following the
distribution
. The
Generalized Nataf transformation
is defined
by:
where the three transformations ,
and
are given by:
where is the cumulative distribution function of the
standard 1-dimensional elliptical distribution with characteristic
generator
and
is the inverse of the
Cholesky factor of
.
The distribution of is the
generic elliptical representative associated to the copula of
. The step
maps this distribution into its
standard representative, following exactly the same algebra as the
normal copula. Thus, in the Generalized Nataf standard space, the
random vector
follows the standard representative
distribution of the copula of the physical random vector
.
If the copula of is normal,
follows
the standard normal distribution with independent components.