# Eﬃcient estimation of conditional covariance matrices

for dimension reduction
Maikol Sol´ıs Chac´on Jean-Michel Loubes Cl´ement Marteau
Institut de Math´ematiques de Toulouse
Universit´e Paul Sabatier
44
e
Journ´ees de Statistiques
Bruxelles
22 August 2012
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 1 / 23
Outline
1
Introduction
2
Methodology
3
4
Main results
5
Summary
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 2 / 23
Introduction
Outline
1
Introduction
2
Methodology
3
4
Main results
5
Summary
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 3 / 23
Introduction
Background: Sliced Inverse Regression [Li, 1991]
Consider the problem
Y = ϕ(X) + ,
where X ∈ R
p
, Y ∈ R and E[] = 0.
Find β’s such that
Y = φ(β

1
X, . . . , β

K
X, ),
where the β’s are unknown vectors in R
p
, is independent of X and
φ is an arbitrary function in R
K+1
. K p.
The eigenvectors of Cov
_
E[X|Y]
_
spans the same subspace that the
β’s (eﬀective dimension reduction directions or e.d.r.d’s).
The objective is to estimate Cov
_
E[X|Y]
_
.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 4 / 23
Introduction
Background: Sliced Inverse Regression [Li, 1991]
Consider the problem
Y = ϕ(X) + ,
where X ∈ R
p
, Y ∈ R and E[] = 0.
Find β’s such that
Y = φ(β

1
X, . . . , β

K
X, ),
where the β’s are unknown vectors in R
p
, is independent of X and
φ is an arbitrary function in R
K+1
. K p.
The eigenvectors of Cov
_
E[X|Y]
_
spans the same subspace that the
β’s (eﬀective dimension reduction directions or e.d.r.d’s).
The objective is to estimate Cov
_
E[X|Y]
_
.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 4 / 23
Introduction
Background: Sliced Inverse Regression [Li, 1991]
Consider the problem
Y = ϕ(X) + ,
where X ∈ R
p
, Y ∈ R and E[] = 0.
Find β’s such that
Y = φ(β

1
X, . . . , β

K
X, ),
where the β’s are unknown vectors in R
p
, is independent of X and
φ is an arbitrary function in R
K+1
. K p.
The eigenvectors of Cov
_
E[X|Y]
_
spans the same subspace that the
β’s (eﬀective dimension reduction directions or e.d.r.d’s).
The objective is to estimate Cov
_
E[X|Y]
_
.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 4 / 23
Introduction
Background: Sliced Inverse Regression [Li, 1991]
Consider the problem
Y = ϕ(X) + ,
where X ∈ R
p
, Y ∈ R and E[] = 0.
Find β’s such that
Y = φ(β

1
X, . . . , β

K
X, ),
where the β’s are unknown vectors in R
p
, is independent of X and
φ is an arbitrary function in R
K+1
. K p.
The eigenvectors of Cov
_
E[X|Y]
_
spans the same subspace that the
β’s (eﬀective dimension reduction directions or e.d.r.d’s).
The objective is to estimate Cov
_
E[X|Y]
_
.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 4 / 23
Introduction
Previous work
Estimators for Cov
_
E[X|Y]
_
and the e.d.r.d’s can be found using
Kernel estimators, [Zhu & Fang, 1996] and
[Ferr´e & Yao, 2003,2005].
A combination of the nearest neighbor and SIR, [Hsing 1999].
Assumption that E[X|Y] has some parametric form,
[Bura & Cook, 2001].
K-means, [Setodji & Cook,2004].
Transformation of SIR to least square form,
[Cook & Ni, 2005].
etc.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 5 / 23
Introduction
Our idea to estimate Cov
_
E[X|Y]
_
The objective is to estimate directly Cov
_
E[X|Y]
_
, using a plug-in
method in a semiparametric framework.
We aim to estimate E
_
E[X|Y] E[X|Y]

_
(studied by [Da Veiga & Gamboa, 2012] and [Laurent, 1996]).
We will also focus on the eﬃcient estimation of
E
_
E[X|Y] E[X|Y]

_
.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 6 / 23
Introduction
Our idea to estimate Cov
_
E[X|Y]
_
The objective is to estimate directly Cov
_
E[X|Y]
_
, using a plug-in
method in a semiparametric framework.
We aim to estimate E
_
E[X|Y] E[X|Y]

_
(studied by [Da Veiga & Gamboa, 2012] and [Laurent, 1996]).
We will also focus on the eﬃcient estimation of
E
_
E[X|Y] E[X|Y]

_
.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 6 / 23
Introduction
Our idea to estimate Cov
_
E[X|Y]
_
The objective is to estimate directly Cov
_
E[X|Y]
_
, using a plug-in
method in a semiparametric framework.
We aim to estimate E
_
E[X|Y] E[X|Y]

_
(studied by [Da Veiga & Gamboa, 2012] and [Laurent, 1996]).
We will also focus on the eﬃcient estimation of
E
_
E[X|Y] E[X|Y]

_
.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 6 / 23
Methodology
Outline
1
Introduction
2
Methodology
3
4
Main results
5
Summary
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 7 / 23
Methodology
Preliminaries
1
Let f (x
i
, x
j
, y) be the joint distribution of (X
i
, X
j
, Y), then deﬁne
T
ij
(f ) = E
_
E[X
i
|Y] E[X
j
|Y]

_
.
2
In general, for f ∈ L
2
(dx
i
, dx
j
, dy), deﬁne the non-linear functional
f →T
ij
(f ) with T
ij
(f ) having the form
_ _
_
x
i
f (x
i
, x
j
, y)dx
i
dx
j
_
f (x
i
, x
j
, y)dx
i
dx
j
__
_
x
j
f (x
i
, x
j
, y)dx
i
dx
j
_
f (x
i
, x
j
, y)dx
i
dx
j
_
f (x
i
, x
j
, y)dx
i
dx
j
dy.
3
For an i.i.d sample (X
(k)
i
, X
(k)
j
, Y
(k)
), k = 1, . . . , n, we can build a
preliminary estimator
ˆ
f of f with a subsample n
1
< n.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 8 / 23
Methodology
Preliminaries
1
Let f (x
i
, x
j
, y) be the joint distribution of (X
i
, X
j
, Y), then deﬁne
T
ij
(f ) = E
_
E[X
i
|Y] E[X
j
|Y]

_
.
2
In general, for f ∈ L
2
(dx
i
, dx
j
, dy), deﬁne the non-linear functional
f →T
ij
(f ) with T
ij
(f ) having the form
_ _
_
x
i
f (x
i
, x
j
, y)dx
i
dx
j
_
f (x
i
, x
j
, y)dx
i
dx
j
__
_
x
j
f (x
i
, x
j
, y)dx
i
dx
j
_
f (x
i
, x
j
, y)dx
i
dx
j
_
f (x
i
, x
j
, y)dx
i
dx
j
dy.
3
For an i.i.d sample (X
(k)
i
, X
(k)
j
, Y
(k)
), k = 1, . . . , n, we can build a
preliminary estimator
ˆ
f of f with a subsample n
1
< n.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 8 / 23
Methodology
Preliminaries
1
Let f (x
i
, x
j
, y) be the joint distribution of (X
i
, X
j
, Y), then deﬁne
T
ij
(f ) = E
_
E[X
i
|Y] E[X
j
|Y]

_
.
2
In general, for f ∈ L
2
(dx
i
, dx
j
, dy), deﬁne the non-linear functional
f →T
ij
(f ) with T
ij
(f ) having the form
_ _
_
x
i
f (x
i
, x
j
, y)dx
i
dx
j
_
f (x
i
, x
j
, y)dx
i
dx
j
__
_
x
j
f (x
i
, x
j
, y)dx
i
dx
j
_
f (x
i
, x
j
, y)dx
i
dx
j
_
f (x
i
, x
j
, y)dx
i
dx
j
dy.
3
For an i.i.d sample (X
(k)
i
, X
(k)
j
, Y
(k)
), k = 1, . . . , n, we can build a
preliminary estimator
ˆ
f of f with a subsample n
1
< n.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 8 / 23
Methodology
Taylor’s development
Deﬁne the auxiliar function F : [0, 1] →R as
F(u) = T
ij
(uf + (1 −u)
ˆ
f ) with u ∈ [0, 1].
We have F(1) = F(0) + F

(0) +
1
2
F

(0) +
1
6
F

(ξ)(1 −ξ)
3
for some
ξ ∈ [0, 1].
Notice that F(1) = T
ij
(f ) and F(0) = T
ij
(
ˆ
f ).
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 9 / 23
Methodology
Taylor’s development
Deﬁne the auxiliar function F : [0, 1] →R as
F(u) = T
ij
(uf + (1 −u)
ˆ
f ) with u ∈ [0, 1].
We have F(1) = F(0) + F

(0) +
1
2
F

(0) +
1
6
F

(ξ)(1 −ξ)
3
for some
ξ ∈ [0, 1].
Notice that F(1) = T
ij
(f ) and F(0) = T
ij
(
ˆ
f ).
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 9 / 23
Methodology
Taylor’s development
Deﬁne the auxiliar function F : [0, 1] →R as
F(u) = T
ij
(uf + (1 −u)
ˆ
f ) with u ∈ [0, 1].
We have F(1) = F(0) + F

(0) +
1
2
F

(0) +
1
6
F

(ξ)(1 −ξ)
3
for some
ξ ∈ [0, 1].
Notice that F(1) = T
ij
(f ) and F(0) = T
ij
(
ˆ
f ).
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 9 / 23
Methodology
Taylor’s development
We obtain,
T
ij
(f ) =
_
H
1
(
ˆ
f , x
i
, x
j
, y)f (x
i
, x
j
, y)dx
i
dx
j
dy
. ¸¸ .
Linear Functional (LF)
+
_
H
2
(
ˆ
f , x
i 1
, x
j 2
, y)f (x
i 1
, x
j 1
, y)f (x
i 2
, x
j 2
, y)dx
i 1
dx
j 1
dx
i 2
dx
j 2
dy
. ¸¸ .
+ Γ
n
.¸¸.
Error
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 10 / 23
Methodology
Estimating (LF) and (QF)
The linear part is direct.
Deﬁne n
2
= n −n
1
. We estimate (LF) by
1
n
2
n
2

k=1
H
1
_
ˆ
f , X
(k)
i
, X
(k)
j
, Y
(k)
_
.
Core of the work: Main issue.
To estimate (QF), we will build an estimator of
θ(f ) =
_
η(x
i 1
, x
j 2
, y)f (x
i 1
, x
j 1
, y)f (x
i 2
, x
j 2
, y)dx
i 1
dx
j 1
dx
i 2
dx
j 2
dy
where η : R
3
→R is a bounded function.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 11 / 23
Methodology
Estimating (LF) and (QF)
The linear part is direct.
Deﬁne n
2
= n −n
1
. We estimate (LF) by
1
n
2
n
2

k=1
H
1
_
ˆ
f , X
(k)
i
, X
(k)
j
, Y
(k)
_
.
Core of the work: Main issue.
To estimate (QF), we will build an estimator of
θ(f ) =
_
η(x
i 1
, x
j 2
, y)f (x
i 1
, x
j 1
, y)f (x
i 2
, x
j 2
, y)dx
i 1
dx
j 1
dx
i 2
dx
j 2
dy
where η : R
3
→R is a bounded function.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 11 / 23
Outline
1
Introduction
2
Methodology
3
4
Main results
5
Summary
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 12 / 23
Estimation of θ(f ): Projection method.
Assumption
S
M
n
f −f
2
2
→0 where S
M
n
f =

l ∈M
n
a
l
p
l
(p
l
is a basis of
L
2
(dx
i
, dx
j
, dy)).
_
sup
l / ∈M
n
|c
l
|
2
_
2
≈ |M
n
| /n
2
for c
l
a ﬁxed sequence.
We build an estimator of θ using a projection scheme with Bias equal
to

_
(S
M
n
f (x
i 1
, x
j 1
, y)−f (x
i 1
, x
j 1
, y))(S
M
n
f (x
i 2
, x
j 2
, y)−f (x
i 2
, x
j 2
, y))
η(x
i 1
, x
j 2
, y)dx
i 1
dx
j 1
dx
i 2
dx
j 2
dy
(AMSE) E
_
ˆ
θ
n
−θ
_
2
= O
_
1
n
_
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 13 / 23
Estimation of θ(f ): Projection method.
Assumption
S
M
n
f −f
2
2
→0 where S
M
n
f =

l ∈M
n
a
l
p
l
(p
l
is a basis of
L
2
(dx
i
, dx
j
, dy)).
_
sup
l / ∈M
n
|c
l
|
2
_
2
≈ |M
n
| /n
2
for c
l
a ﬁxed sequence.
We build an estimator of θ using a projection scheme with Bias equal
to

_
(S
M
n
f (x
i 1
, x
j 1
, y)−f (x
i 1
, x
j 1
, y))(S
M
n
f (x
i 2
, x
j 2
, y)−f (x
i 2
, x
j 2
, y))
η(x
i 1
, x
j 2
, y)dx
i 1
dx
j 1
dx
i 2
dx
j 2
dy
(AMSE) E
_
ˆ
θ
n
−θ
_
2
= O
_
1
n
_
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 13 / 23
Estimation of θ(f ): Projection method.
Assumption
S
M
n
f −f
2
2
→0 where S
M
n
f =

l ∈M
n
a
l
p
l
(p
l
is a basis of
L
2
(dx
i
, dx
j
, dy)).
_
sup
l / ∈M
n
|c
l
|
2
_
2
≈ |M
n
| /n
2
for c
l
a ﬁxed sequence.
We build an estimator of θ using a projection scheme with Bias equal
to

_
(S
M
n
f (x
i 1
, x
j 1
, y)−f (x
i 1
, x
j 1
, y))(S
M
n
f (x
i 2
, x
j 2
, y)−f (x
i 2
, x
j 2
, y))
η(x
i 1
, x
j 2
, y)dx
i 1
dx
j 1
dx
i 2
dx
j 2
dy
(AMSE) E
_
ˆ
θ
n
−θ
_
2
= O
_
1
n
_
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 13 / 23
Main results
Outline
1
Introduction
2
Methodology
3
4
Main results
5
Summary
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 14 / 23
Main results
Putting (LF) and (QF) together
´
T
(n)
ij
=
1
n
2
n
2

k=1
H
1
(
ˆ
f , X
(k)
i
, X
(k)
j
, Y
(k)
)+Estimator of (QF)
The estimator of (QF) has two sums which depends on the basis p
l
chosen for the projection.
The term Γ
n
is negligible.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 15 / 23
Main results
Asymptotic normality of
´
T
(n)
ij
Theorem
Under some technical assumptions and |M
n
| /n →0 when n →∞. We
have:

n
_
´
T
(n)
ij
−T
ij
(f )
_
D
−→N (0, C
ij
(f )) ,
and
lim
n→∞
nE
_
´
T
(n)
ij
−T
ij
(f )
_
2
= C
ij
(f ),
where
C
ij
(f ) = Var
_
H
1
(f , X
i
, X
j
, Y)
_
Conclusion: There is asymptotic normality.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 16 / 23
Main results
Semi-parametric Cram´er-Rao bound
Theorem
Under the same conditions as before, for any estimator
´
T
(n)
ij
of T
ij
(f ) and
every family {V
r
(f )}
r >0
of neighborhoods of f we have
inf
{V
r
(f )}
r >0
liminf
n→∞
sup
f ∈V
r
(f )
nE
_
´
T
(n)
ij
−T
ij
(f )
_
2
≥ C
ij
(f ).
Conclusion: The estimator is eﬃcient.
Generalization via half-vectorization to matrices: T(f ) = (T
ij
(f ))
p×p
and
H
1
(f ) = (H
1
(f , x
i
, x
j
, y))
p×p
.

n vech
_
´
T
(n)
−T(f )
_
D
−→N (0, C(f )) ,
C(f ) = Cov
_
vech(H
1
(f ))
_
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 17 / 23
Main results
Semi-parametric Cram´er-Rao bound
Theorem
Under the same conditions as before, for any estimator
´
T
(n)
ij
of T
ij
(f ) and
every family {V
r
(f )}
r >0
of neighborhoods of f we have
inf
{V
r
(f )}
r >0
liminf
n→∞
sup
f ∈V
r
(f )
nE
_
´
T
(n)
ij
−T
ij
(f )
_
2
≥ C
ij
(f ).
Conclusion: The estimator is eﬃcient.
Generalization via half-vectorization to matrices: T(f ) = (T
ij
(f ))
p×p
and
H
1
(f ) = (H
1
(f , x
i
, x
j
, y))
p×p
.

n vech
_
´
T
(n)
−T(f )
_
D
−→N (0, C(f )) ,
C(f ) = Cov
_
vech(H
1
(f ))
_
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 17 / 23
Summary
Outline
1
Introduction
2
Methodology
3
4
Main results
5
Summary
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 18 / 23
Summary
Summary
We use a plug-in method used by
[Laurent, 1996, Da Veiga & Gamboa, 2012] to ﬁnd Cov
_
E[X|Y]
_
.
Taylor’s development on E
_
E[X|Y] E[X|Y]

_
(semi-parametric
framework)
Projection method to estimate the non-linear part.
Asymptotically normal and eﬃcient.
Generalization of the asymptotic normality to the whole matrix.
Non-adaptative: the method depends directly on the regularity of f .
Future work
Numerical results for this method (In progress).
Analyze rates of convergence exploring other techniques like kernel
estimators.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 19 / 23
Summary
Summary
We use a plug-in method used by
[Laurent, 1996, Da Veiga & Gamboa, 2012] to ﬁnd Cov
_
E[X|Y]
_
.
Taylor’s development on E
_
E[X|Y] E[X|Y]

_
(semi-parametric
framework)
Projection method to estimate the non-linear part.
Asymptotically normal and eﬃcient.
Generalization of the asymptotic normality to the whole matrix.
Non-adaptative: the method depends directly on the regularity of f .
Future work
Numerical results for this method (In progress).
Analyze rates of convergence exploring other techniques like kernel
estimators.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 19 / 23
Summary
Summary
We use a plug-in method used by
[Laurent, 1996, Da Veiga & Gamboa, 2012] to ﬁnd Cov
_
E[X|Y]
_
.
Taylor’s development on E
_
E[X|Y] E[X|Y]

_
(semi-parametric
framework)
Projection method to estimate the non-linear part.
Asymptotically normal and eﬃcient.
Generalization of the asymptotic normality to the whole matrix.
Non-adaptative: the method depends directly on the regularity of f .
Future work
Numerical results for this method (In progress).
Analyze rates of convergence exploring other techniques like kernel
estimators.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 19 / 23
Summary
Summary
We use a plug-in method used by
[Laurent, 1996, Da Veiga & Gamboa, 2012] to ﬁnd Cov
_
E[X|Y]
_
.
Taylor’s development on E
_
E[X|Y] E[X|Y]

_
(semi-parametric
framework)
Projection method to estimate the non-linear part.
Asymptotically normal and eﬃcient.
Generalization of the asymptotic normality to the whole matrix.
Non-adaptative: the method depends directly on the regularity of f .
Future work
Numerical results for this method (In progress).
Analyze rates of convergence exploring other techniques like kernel
estimators.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 19 / 23
Summary
Summary
We use a plug-in method used by
[Laurent, 1996, Da Veiga & Gamboa, 2012] to ﬁnd Cov
_
E[X|Y]
_
.
Taylor’s development on E
_
E[X|Y] E[X|Y]

_
(semi-parametric
framework)
Projection method to estimate the non-linear part.
Asymptotically normal and eﬃcient.
Generalization of the asymptotic normality to the whole matrix.
Non-adaptative: the method depends directly on the regularity of f .
Future work
Numerical results for this method (In progress).
Analyze rates of convergence exploring other techniques like kernel
estimators.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 19 / 23
Summary
Summary
We use a plug-in method used by
[Laurent, 1996, Da Veiga & Gamboa, 2012] to ﬁnd Cov
_
E[X|Y]
_
.
Taylor’s development on E
_
E[X|Y] E[X|Y]

_
(semi-parametric
framework)
Projection method to estimate the non-linear part.
Asymptotically normal and eﬃcient.
Generalization of the asymptotic normality to the whole matrix.
Non-adaptative: the method depends directly on the regularity of f .
Future work
Numerical results for this method (In progress).
Analyze rates of convergence exploring other techniques like kernel
estimators.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 19 / 23
Summary
Summary
We use a plug-in method used by
[Laurent, 1996, Da Veiga & Gamboa, 2012] to ﬁnd Cov
_
E[X|Y]
_
.
Taylor’s development on E
_
E[X|Y] E[X|Y]

_
(semi-parametric
framework)
Projection method to estimate the non-linear part.
Asymptotically normal and eﬃcient.
Generalization of the asymptotic normality to the whole matrix.
Non-adaptative: the method depends directly on the regularity of f .
Future work
Numerical results for this method (In progress).
Analyze rates of convergence exploring other techniques like kernel
estimators.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 19 / 23
Summary
Summary
We use a plug-in method used by
[Laurent, 1996, Da Veiga & Gamboa, 2012] to ﬁnd Cov
_
E[X|Y]
_
.
Taylor’s development on E
_
E[X|Y] E[X|Y]

_
(semi-parametric
framework)
Projection method to estimate the non-linear part.
Asymptotically normal and eﬃcient.
Generalization of the asymptotic normality to the whole matrix.
Non-adaptative: the method depends directly on the regularity of f .
Future work
Numerical results for this method (In progress).
Analyze rates of convergence exploring other techniques like kernel
estimators.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 19 / 23
Summary
Summary
We use a plug-in method used by
[Laurent, 1996, Da Veiga & Gamboa, 2012] to ﬁnd Cov
_
E[X|Y]
_
.
Taylor’s development on E
_
E[X|Y] E[X|Y]

_
(semi-parametric
framework)
Projection method to estimate the non-linear part.
Asymptotically normal and eﬃcient.
Generalization of the asymptotic normality to the whole matrix.
Non-adaptative: the method depends directly on the regularity of f .
Future work
Numerical results for this method (In progress).
Analyze rates of convergence exploring other techniques like kernel
estimators.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 19 / 23
Thank you
References
Da Veiga, S. & Gamboa, F. (2012).
Eﬃcient estimation of sensitivity indices.
Arxiv Preprint ArXiv:1203.2899.
Laurent, B. (1996).
Eﬃcient estimation of integral functionals of a density.
The Annals of Statistics, 24(2), 659–681.
Li, K. C. (1991).
Sliced inverse regression for dimension reduction.
Journal of the American Statistical Association, 86(414), 316–327.
Sol´ıs Chac´on, M., Loubes, J. M., Marteau, C., & Da Veiga, S. (2011).
Eﬃcient estimation of conditional covariance matrices for dimension
reduction.
Summary
Hypothesis
Let(p
l
(x
i
, x
j
, y))
l ∈D
an orthonormal basis of L
2
(dx
i
dx
j
dy) (D
countable).
E =
_

l ∈D
a
l
p
l
: (a
l
)
l ∈D
such that

l ∈D
¸
¸
¸
¸
a
l
c
l
¸
¸
¸
¸
2
< 1
_
⊂ L
2
(dx
i
dx
j
dy)
where (c
l
)
l ∈D
is a ﬁxed sequence.
f ∈ E.
Let (M
n
)
n≥1
a subset sequence of D. For any n there is M
n
such that
M
n
⊂ D. Denote by |M
n
| the cardinal of M
n
.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 21 / 23
Summary
Assumptions
1
For any n ≥ 1cmmere is M
n
⊂ D such that
_
sup
l / ∈M
n
|c
l
|
2
_
2
≈ |M
n
| /n
2
. Moreover, ∀f ∈ L
2
(dxdydz),
_
(S
M
n
f −f )
2
dxdydz →0 when n →0, where S
M
n
f =

l ∈M
n
a
l
p
l
.
2
supp f ⊂ [d
1
, b
1
] ×[d
2
, b
2
] ×[d
3
, b
3
] and ∀(x, y, z) ∈ supp f ,
0 < α ≤ f (x, y, z) ≤ β with α, β ∈ R.
3
We can build
ˆ
f , such that > 0,
∀(x, y, z) ∈ supp f , 0 < α − ≤
ˆ
f (x, y, z) ≤ β +
y ∀ 2 ≤ q ≤ +∞, ∀l ∈ N

, E
f
_

f −f
_
_
l
q
≤ C(q, l )n
−l λ
1
for λ > 1/6 and a constant C(q, l ) no depending of f ∈ E.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 22 / 23
Summary
Appendix: Estimator (LF) and (QF)
´
T
(n)
ij
=
1
n
2
n
2

k=1
H
1
(
ˆ
f , X
(k)
i
, X
(k)
j
, Y
(k)
)
+
1
n
2
(n
2
−1)

l ∈M
n
2

k=k

=1
p
l
_
X
(k)
i
, X
(k)
j
, Y
(k)
_
_
p
l
_
x
i
, x
j
, Y
(k

)
_
H
3
_
ˆ
f , x
i
, x
j
, X
(k

)
i
, X
(k

)
j
, Y
(k

)
_
dx
i
dx
j

1
n
2
(n
2
−1)

l ,l

∈M
n
2

k=k

=1
p
l
_
X
(k)
i
, X
(k)
j
, Y
(k)
_
p
l

_
X
(k

)
i
, X
(k

)
j
, Y
(k

)
_
_
p
l
(x
i 1
, x
j 1
, y)p
l
(x
i 2
, x
j 2
, y)H
2
_
ˆ
f , x
i 1
, x
j 2
, y)dx
i 1
dx
j 1
dx
i 2
dx
j 2
dy.
where H
3
(f , x
i 1
, x
j 1
, x
i 2
, x
j 2
, y) = H
2
(f , x
i 1
, x
j 2
, y) + H
2
(f , x
i 2
, x
j 1
, y) and
n
2
= n −n
1
.
Sol´ıs Chac´on, Loubes, Marteau (IMT) Bruxelles, 22 August 2012 23 / 23