Professional Documents
Culture Documents
A Kolmogorov-Smirnov Type Test For Positive Quadrant Dependence
A Kolmogorov-Smirnov Type Test For Positive Quadrant Dependence
/ La Revue Canadienne de Statistique, Vol. 33, No. 3, Dependence Modelling: Statistical Theory and Applications in Finance and Insurance (Sep., 2005), pp. 415-427 Published by: Statistical Society of Canada Stable URL: http://www.jstor.org/stable/25046188 . Accessed: 09/05/2011 12:31
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=ssc. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Statistical Society of Canada is collaborating with JSTOR to digitize, preserve and extend access to The Canadian Journal of Statistics / La Revue Canadienne de Statistique.
http://www.jstor.org
Hie Canadian Journal of Statistics Vol. 33, No. 3,2005, Pages 415-427 La revue canadienne de statistique
415
secondary 60E15,62P05.
Kolmogorov-Smirnov type of test of the complete set of
a consistent,
restrictions that relate to the copula representation of positive quadrant dependence. For such a test, he pro poses and justifies inference relying on a simulation-based multiplier method and a bootstrap method. He also explores the finite-sample behaviour of both methods with Monte Carlo experiments. A first empirical illustration is given for American insurance claim data. A second one examines the presence of positive
quadrant dependence in Ufe expectancies at birth of males and females across countries.
pour
de
la d?pendance
positive
positive
de par quadrant
par quadrant
l'ensemble sous forme des complet de copule.
type Kolmogorov-Smirnov
la d?pendance
Pour un tel test, il propose et justifie une inference reposant sur une m?thode simul?e ?multiplicateurs et une m?thode de bootstrap. Il explore aussi le comportement ? distance finie des deux m?thodes ? l'aide d'exp?riences deMonte-Carlo. Une premi?re illustration empirique est faite pour des donn?es am?ricaines
de sinistres en assurance. de vie des Une seconde et femmes examine la pr?sence de d?pendance positive par quadrant entre les esp?rances hommes par pays.
1. INTRODUCTION
The concept of positive quadrant dependence (PQD) was introduced by Lehmann (1966). Two random variables are said to be PQD when the probability that they are simultaneously large (or small) is at least as great as itwould be were they independent. in finance, insurance and risk management has emphasized the importance of see, e.g., Dhaene & Goovaerts (1996), Denuit, Dhaene & Ribas (2001), Embrechts, Mc PQD; Neil & Straumann (2000). One advantage of this dependence structure is that it allows the risk manager to compare directly the sum of PQD random variables with the corresponding sum under the independence assumption. The comparison is in the sense of different stochastic orderings Inferring that two claims are expressing the common preferences of rational decision makers. no matter what the strength of this dependence is, immediately allows one to conclude PQD, the underestimation of most insurance premiums involving a portfolio of these two claims if the independence assumption ismade instead. In a financial setting, the same holds true, but for risk measures and derivative prices re lated to a portfolio of two PQD financial assets. We refer the reader to Denuit 8c Scaillet (2004) for an extensive discussion of examples of the application of PQD in finance and actuarial sci ences. Implications of PQD are also found in reliability theory and several other fields; see, e.g., Levy (1992), Shaked & Shanthikumar (1994), Drouef-Mari & Kotz (2001) or Lai & Xie (2003). Formally, two random variables X and Y are said to be PQD if, for all (x, y) R2, Recent work
P(* < a, Y < y) > P(X < x)P{Y < y). (1)
Of course, (1) is equivalent to
P(X >x,Y>y)>
(2)
416
SCAILLET
which enjoys a similar interpretation (with "small" replaced by "large"). Note further that nega tive quadrant dependence (NQD) is defined analogously if we substitute < for > in (1). (l)-(2), PQD appears as a comparison of the joint distribution of (X, Y) to Considering that of (X, Y)x, where (X, Y)1- denotes an independent version of the random vector (X, Y), and (X, Y)1- has independent i.e., (X, Y) and (X, Y)L have identical univariate marginals, It can thus be considered as a special case of comparisons of pairs of bivariate components. distributions with identical marginals in terms of stochastic dominance. Clearly, X and Y are PQD if, and only if, a(X) and b(Y) are PQD for any strictly increasing functions a and b. This indicates that PQD is a property of the underlying copula, and is not influenced by the marginals. In fact, the inequality (1) can be written in terms of the copula C of the two random variables, since (1) is equivalent to the condition that, for all (i?, v) G [0,1]2, C(u,v) Recall that by definition >C?(u,v)=uv. (3) is such that
2005
Following Deheuvels
417
Cn(u,v)
-f]l{Fn(Xi) n r?f
(u,v) e [0,1]2,
where Fn and Gn are the empirical cumulative distribution functions computed from X\,..., Xn and Y?,..., Yn, respectively. Observe that Cn is actually a function of the ranks of the observa tions since nFn(Xi) is the rank of Xi among X\,..., Xn, and likewise for nGn(1?). Let Dn(u, v) = uv - Cn(u, v), D(u, v) = uv C(u, v) and denote by ???([0, l]2) the set of all locally bounded real functions on [0, l]2. LEMMA 1. = Assume that the copula derivatives function C has continuous partial ? = Then^(Dn C2(u,v). D) converges weakly in ???([0, l]2), whose covariance function is
ttG{u,v,u',vf)
=
=
E{Gc(u,v)Gc(uV)}
??(u, v, u\ v') Ci(u', t/)?l(u, v, v!, 1) C2(u', v')Sl(u, Ci(w, v){??(w, 1, v!, vf) Ci(v!, i/)?l(u, 1, t*', 1) v, u', t/) C2(ii, v){0(l, C2(i?', t/)fi(ti, 1,1, t/)} -Ci(u,,t/)il(l,t;,ti,,l)-C2(ti/,i/)ft(l,v,l,t/)}, v, 1, t/)
G [0,1], wftere ?2(i?,u,i?',i;') /or a/Z i?,i?',v,vf x A y = min(x, y). This lemma follows
C(u
Au\v
A v') ? C(u,v)C(u\v')
and
from the weak convergence properties of the empirical copula process Its proof, given in the Appendix, relies on Theorem 4 in Fermanian, Radulovic & C). V (?n (2004), which shows that the result stated by van der Vaart & Wellner (1996, p. 389) Wegkamp see also Stute (1984) or G?nssler & holds true in a larger space under weaker assumptions; Stute (1987) for weak convergence of the same process in the Skorokhod space ?>([0, l]2). ? that uniform almost sure convergence is a by-product of this type of weak convergence. In the special case C = CL, the weak convergence of y/? (Cn C) towards a completely-tucked Brownian sheet was originally given by Deheuvels (1981), who developed a Kolmogorov Note Smirnov Since we wish test for independence. to test for PQD, namely Ho Hi we consider the test statistic : C(u, : C(u, v) > uv for all (u, v) ? [0, l]2, for some (u, v) G [0, l]2,
v) < uv
Sn = V?
and a test based on the decision will be discussed The following later. result characterizes
rule "reject Ho
if Sn >
c," where
S =
Gc(u,y).
Proposition i) IfHo
418
?) If Wo isfalse, then
SCAILLET
n-?oo
lim P(rejectHo) = 1.
The first part of the result provides a random variable that dominates the limiting random variable corresponding to the test statistic under the null hypothesis. The inequality tells us that the test will never reject more often than a(c) when the null hypothesis is satisfied. Furthermore, the probability of rejection will asymptotically be exactly a(c) when C = CL. The first part also implies that if one could find a c to set the a(c) to some desired probability level (say the conventional 0.05 or 0.01), then this would be the significance level for composite null hypothe ses in the sense described by Lehmann (1986). The second part of the result indicates that the test is capable of detecting any violation of the full set of restrictions of the null hypothesis. Of course, in order to make the result operational, we need to find an appropriate critical value c. Since the distribution of the test statistic depends on the underlying copula, this is not an easy task. Indeed, recall that the null hypothesis is not independence. Therefore, we cannot i.e., draw from the independent copula. directly simulate under the independence hypothesis; Such a procedure would not reflect the dependence of the distribution of the test statistic on the underlying copula. Hereafter we rely on two different methods to simulate P-values.
3. SIMULATING P-VALUES
3.1. Multiplier method.
In this section, we use a simulation-based method that exploits themultiplier central limit theory discussed in Section 2.9 of van der Vaart & Wellner (1996). See Barrett & Donald (2003) for a similar use in the context of a stochastic dominance test, and Hansen (1996), Glidden (1999), Guay & Scaillet (2003) for other applications. The idea is to rely on artificial pseudo-random numbers to simulate a process that is iden tical but (asymptotically) Un denote a sequence of independent of Gc> To do this, let Ui,..., are independent of the data independent and identically distributed N(0,1) random variables that sample. Since Gc(ti, where Be v) = Bc(ti, v) Ci(u, v)Bc(u, 1) C2(ti, v)Bc(l, t/), is easily
is a tight Brownian
generated from
GCw(ti,t;)
x x
c2,n(u, W)-L?Ui
where
= 1,2. Consistent estimates are easily c?>n(t?, v) is a consistent estimate of Ci(u, v) i obtained from smoothed versions of the empirical copula process. For example, one can use the nonparametric estimators
ci,n(t*,t;)
and
= ^Hn{F-\u),G-\v)}lfn{F-\u)}
= C2,?(?,t;) ^Hn{F-l{u),G-l{v)}lgn{G-\v)},
2005
419
where estimates of the joint cumulative distribution function H and themarginal densities / and g are obtained from kernel based estimators Hn, fn and gn, respectively; see Fermanian & Scail let (2003) for details and proofs of asymptotic properties of such estimators. The P-value can be estimated from
1/2,
if
<a
if Ho
is true,
n?fco
ifHo isfalse.
The multiplier method can be justified by showing that the simulated process converges weakly to an identical independent copy of the Gaussian process <Gc. Then an application of the continuous mapping theorem shows that we get a simulated copy of the bounding random variable that appears in Proposition 1. In practice, we use Monte Carlo methods to approximate the probability and a grid to approximate the supremum. The P-value is simply approximated by
1
where Note
R
r=l
the averaging is done on R replications, and 5n,r is computed from a fine grid on [0, l]2. that the replication number and the grid mesh can be chosen tomake the approximations as accurate as one desires given time and computer constraints.
3.2. Bootstrap method. second method relies on the standard bootstrap; see Abadie (2002) and Barrett & Don ald (2003) for similar applications in the context of stochastic dominance tests. An alternative resampling technique could be subsampling, for which similar results can be shown to hold as well; see Linton, Maasoumi & Whang (2005) for its use in stochastic dominance tests. The C* Let (X*, F]*),..., (X*, Y*) be a random sample drawn from the observed pairs of data, and the empirical copula function built from this bootstrap sample. Let us further take S*n and define = y/n sup
u,ve[o,i]
K
Then th? bootstrap method
p(s;>Sn).
is justified by the next statement. 1/2, a test for PQD based on the rule "reject Ho if
Again, we need to rely on Monte Carlo methods to approximate the probability and a grid to approximate the supremum in amanner analogous to that in the previous subsection.
SCAILLET
In this section, we examine the performance of the Kolmogorov-Smirnov type test in small The grid is made of the values (u, v) evenly spaced inside {0.05,0.10,..., samples. 0.95} x estimator of the derivatives of the copula func 0.95}, while the nonparametric {0.05,0.10,..., tion relies on a Gaussian product kernel and the quick standard rule of thumb (Silverman 1986) to select the two individual bandwidths. The replication number R to approximate the P-value is set equal to 1000. In each case, 1000 Monte Carlo simulations are performed, and the rejection rates are computed for the multiplier method and the bootstrap method with respect to the two conventional significance levels of a = 0.05 and a = 0.01. Samples are generated with both margins corresponding to an exponential distribution with a unit parameter. This can be seen as mimicking the behaviour of claim or duration data. Note that the numerical results below remain exactly the same if we use other strictly monotonie continuously differentiable cumula as Gaussian or Student margins tomimic financial returns) and tive distribution functions (such generators. The reason is that both procedures rely keep the same seeds in the pseudo-random on ranks. intrinsically In Table 1, the true copula is the independent copula. Then Proposition 1 suggests that the test should reject the null hypothesis with a frequency close to the chosen nominal significance level. This experiment should give us some idea about the validity of the asymptotic theory and the two in small samples. The values shown in Table 1 indicate methods used to simulate the P-values that the test tends to over-reject with the multiplier method and under-reject with the bootstrap method, but with a rejection rate converging in both cases to the chosen nominal significance
level as n increases.
Table
1: Observed significance level of the test when the underlying copula is the independent copula.
n = 50 .103 .024
n = 50
.024 .003
.023 .002
.021 .004
.014 .007
Table 2 gathers results concerning the power of the testing procedure when the true copula is a Frank copula, a Gaussian copula, or a Farlie-Gumbel-Morgenstern copula inducing NQD (?). These parametric families are often used in actuarial and financial applications, and are (1999). The chosen values of the param easy to simulate; see, e.g., Genest (1987) or Nelsen eter 0 are 6 G {-1, -2, -3} for the Frank copula, 0 G {-.17, -.32, -.46} for the Gaussian for the FGM copula. They match low and mod copula, and 0 G {?.495, ?.945, ?1.395} as exhibited by the corresponding true values of the Kendall tau, erate negative dependencies r G {?.11, ?.21, ?.31}. The sample size is fixed at n = 200. The reported numbers show that both testing procedures
tures.
dependence
struc
5. EMPIRICAL ILLUSTRATIONS
5.1. American insurance claims.
Various
processes in casualty insurance involve correlated pairs of variables. A prominent ex is the loss and allocated loss adjustment expenses (ALAE, for short) on a single claim. ample
2005
Here ALAE
421
are a type of insurance company expense that is specifically attributable to the set tlement of individual claims, such as lawyers' fees and claims investigation expenses. The joint in parametric settings of those two variables has already been examined by Frees & modeling Valdez (1998), and Klugman & Parsa (1999). Table 2: Power of the test for Frank (F),Gaussian (G) and Farlie-Gumbel-Morgenstern
alternatives.
(FGM) copula
T=
-0.11
r =
-0.21
r =
-0.31
a = 0.05 Multiplier method Bootstrap method a = 0.01 Multiplier method Bootstrap method
The data used in these empirical studies were collected by the US Insurance Services Of fice, and comprise general liability claims randomly choosen from late settlement lags. Frees & Valdez (1998) choose the Pareto distribution tomodel themargins, and select Gumbel and Frank copulas (on the basis of a graphical procedure suitable for Archimedean copulas). Both models express PQD by their estimated parameter values. Klugman & Parsa (1999) opt for the Inverse Paralogistic for the losses and for the Inverse Burr for ALAEs. They use the Frank copula. Again, the estimated value of the dependence parameter entails PQD for losses and ALAEs. In the following, we rely on a nonparametric approach to assess PQD. This assessment has many implications in insurance, for example, for the computation of reinsurance premi ums (where the sharing of expenses between the ceding company and the reinsurer has to be decided on) and for the determination of the expense level for a given loss level (for reserving an appropriate amount to cover future settlement expenses). We refer to Denuit & Scaillet (2004) for further discussion and the practical implications on the design of reinsurance treaties when PQD is present. The data consist in n = 1466 uncensored observed values of the pair (LOSS,ALAE). The grid, the number of replications, the bivariate kernel and the bandwidths are chosen as in the = 1.000 = 1.000 previous Monte Carlo experiments. We find pn (multiplier method) and pn = which means that we cannot reject PQD. -0.0356, (bootstrap method) for Sn 5.2. Life expectancies at birth. at birth of
This second empirical illustration aims to detect a PQD behaviour in life expectancies males and females across 225 different countries. These data are available at
http://www.odci.gov/cia/publications/factbook/.
A slightly different type of data (life expectancy on total population versus difference be tween life expectancy of males and females) has been examined in Amblard & Girard (2005) in the context of semiparametric estimation of bivariate copulas under a PQD assumption. The grid, the number of replications, the bivariate kernel and the bandwidths are again chosen as in the previous Monte Carlo experiments. We find pn = 1.000 (multiplier method) and p* = 1.000 = which means that we cannot reject PQD. -0.0208, (bootstrap method) for Sn
422
SCAILLET
V(tii,
...,ud)e in short) if
[0, l]d,
(4)
(PUOD,
C(tii,...,ti<f)>C?(tii,...,ttd)
V(ui,...,^)G[0,l]d,
(5)
where C denotes the survival copula associated with C; see Nelsen (1999). Of course, (4) and (5) are no more equivalent when d > 3. When (4) and (5) simultaneously hold, then Y is said to be positively orthant dependent (POD, in short). The extension of the testing procedure in the PLOD case is immediate. the empirical copula process converges weakly to d Gc(tii, - -, Ud) = Bc(wi, - -, ud) ]T Ci(ui,...,
?=i
ud)Bc{l,...,
1, uh 1,...,
1),
where Be E{Bc(ui,...,
udAud)-C(uu
the aforementioned
{C?(uu...,ud)-Cn(ui,...,ud)}.
[0,l]
The PUOD case is more delicate When d = 3, we have C (ui,u2,U2>) = ui +u2 + C(l -
to handle. One could rely on the link between C and C. 2 + C(l ? i?2, 1 -1?3) C(l ?xi,1 ix2)+ C(l -1*1,1-^3) wi, 1 u2,1 u3)
4-U3
2005
423
(see Georges et al. 2001 for a translation formula in the general case) and build the estimator C n obtained from substituting Cn for C. The weak convergence ofthe empirical copula process can then be invoked again to lead to the conclusion that y/? (C n ? C ) has a Gaussian limit. This means that the properties of a testing procedure based on
Sn = y/?
should be similar.
sup
ui,...,ud6[0,l]
{C
(uiJ...,ud)-Cn(ui,...,Ud)}
Finally to derive a test for POD, one could rely on a test statistic equal to the supremum of the bivariate vector made of V?{C-L(uu...,ud)-Cn(uu...1Ud)} and parallel the previous developments. and y/n{C (uu . ..,ud) Cn(uu.. .,ud)},
APPENDIX
All A.l. limits are taken as n -> co. Proof of Lemma 1.
Under continuous differentiability ofthe copula function, Theorem 4 of Fermanian, Radulovic & (2004) states that the empirical copula process converges weakly in ???([0, l]2) to Wegkamp
wards
Gc(u, where Be
v)
Bc(u,
v)
Ci(u,
v)Bc(u,
1)
is a tight Brownian
v')
for each i?, u'yv, v' G [0,1]. This yields the result, upon computation A.2. Proof of Proposition 1.
ofthe
Proof of Part (i): From the definitions (tz, v) G [0, l]2, we obtain Sn < <
v)
<
0 for all
sup \fn{Dn(u,v)
u,v
D(u,v)} D(u,v)}.
+ sup y/nD(u,v)
u,v
supy/n{Dn(u,v)
u,v
Hence ofS.
convergence
of y/? (Dn ? D)
Proof of Part (ii): If the alternative is true, then there is some (u, v), say (?, v) G [0, l]2, for which D(U, v) = 5 > 0. Then the result follows using the inequality Sn > y/?Dn(?, v) and Theorem 4 of Fermanian, Radulovic & Wegkamp (2004). A3. Proof of Proposition!.
Let us write
GCn (u, v)
4= ? -
C(ti, t;)]
424
SCAILLET
c2>n(u,v)^= ???
[l{Gn(Yi) <v}-
v],
{Cn(u,v)-C(u,v)}?Y,Ui.
First Wegkamp consider the last term. Note that Theorem 4 of Fermanian, Radulovic has the property that sample &
sup
u,v [0,l]
v) \Cn(%
C(% v)\?>
0.
we have that condi
and identically
distributed
N(0,1),
U*
sup
[o,i] IV
\Cn(u,v)-C(u,v)\
f I1
~[
-7= y^Ui\>e\
I 1
I J
<
Consequently
1{
Ui4 0,
implies that
{Cn(u, v)
C{u, v)}?= ? *
so that
Ui =*0.
i=l
1
C(u, = v)} ?j=
n
22 Ui => 0> almost surely.
{Cn(ii, v)
||I7||i
E\U\
< oo,
/?OO
= ||I7||2,i /
we deduce
y/P(\U\ >x)dx
< oo
and E max
JO l<i<n
\Ui\/y/? -> 0,
from the multiplier inequalities of Lemma 2.9.1 of van der Vaart & Welltier (1996) that the asymptotic equicontinuity conditions for the three empirical and multiplier processes are equivalent, respectively. This means that the sum of first three terms converge weakly to an independent copy G'c of Gc- As in Theorem 2.9.7 of van der Vaart & Wellner (1996). this leads to the almost sure conditional convergence sup
heBLi
\Euh(Ccn)
Eh(Gc)\
-? 0,
almost surely
where BLi
2005
Now to show the result concerning the asymptotic behaviour of the P-values, let Pn (t) be the cumulative distribution function of the process (conditional on the original sample) generated by sup Gcn The Continuous Mapping Theorem yields sup
u,? [0,l]
Gcn (uiv) =^
u,v
SUP
[0,l]
G>c(u> v)>
where the latter random variable is an independent copy of S. Note that the median of the dis tribution P?(t) of sup G^ is strictly positive and finite. Since Gfc is a Gaussian process indexed by two parameters within the compact set [0,1]2, P? is absolutely continuous (Tsirel'son 1975), while c(a) defined by P{S > c(a)} = a is finite and positive for any a < 1/2 (Proposi tion A.2.7 of van der Vaart & Wellner (1996). The event {pn < a} is equivalent to the event
= = <
lim P{Sn > cn(a)} lim P{Sn > c(a)} + = a, P{S > c(a)}
n?*oo n?>oo
where
the last statement comes from (7), Part (i) of Proposition 1, and c(a) being a continuity point of the distribution of S. On the other hand, Part (ii) of Proposition 1 and c(a) < oo ensure that
n-?oo
lim P(rejecfHol#i)
= 1.
3.
Let C* be the empirical copula associated to the bootstrap sample. Theorem 6 of Fermanian, ? Radulovic & Wegkamp (2004) states that \fn (C* Cn) converges weakly to an independent on the sample in that copy Gq of Gc in probability conditionally
heBL!
0,
given the original sample. Hence we deduce from the Continuous S?=^supG?(ti,t;),
where
the latter random variable is an independent copy of 5, and we can continue as in the of Proposition 2 but using convergence in probability, instead of almost sure convergence, proof to get the final result.
ACKNOWLEDGEMENTS
The author would like to thank Christian Genest and two referees for constructive criticism, as well as Professors E. W. Frees and E. A. Valdez for kindly providing the Loss-ALAE data, which were collected by the US Insurance Services Office (ISO). The author also wishes to express his deep gratitude to Bru no R?millard for his valuable help in designing the simulation-based multiplier method in an appropriate way. He is further grateful to the participants at seminars atULB and CERN as well as at theDeMoSTAFI conference for their comments. The author acknowledges financial support by the Swiss National Science Foundation through theNational Center of Competence: Financial Valuation andRisk Management (NCCR FINRISK). Part of his research was done when he was visiting THEMA and ECARES.
426 REFERENCES
SCAILLET
A. Abadie (2002). Bootstrap tests for distributional treatment effects in instrumental variable models; Jour nal of theAmerican Statistical Association, 97,284-292. C. Amblard & S. Girard (2005). Estimation procedures for a semiparametric family of bivariate copulas.
Journal G. Anderson G. Barrett V. Dardanoni 49-75. of Computational (1996). & S. Donald & A. and Graphical tests Statistics, for stochastic 14,1-15. dominance. Econometrica, 64,1183-1193. Nonparametric (2003).
Consistent
Forcina
(1999).
Inference
R. Davidson & J.-Y. Duelos (2000). Statistical inference for stochastic dominance and for themeasurement of poverty and inequality. Econometrica, 68,1435-1464. P. Deheuvels (1979). La fonction de d?pendance empirique et ses propri?t?s: Un test non param?trique
d'ind?pendance. Bulletin de l'Acad?mie royale de Belgique, Classe des sciences, 65,274-292.
P.Deheuvels
Journal
M. Denuit, J. Dhaene & C. Ribas (2001). Does positive dependence between individual risks increase
stop-loss premiums? Insurance: Mathematics and Economics, 28,305-308.
M. Denuit & O. Scaillet (2004). Nonparametric tests for positive quadrant dependence. Journal ofFinancial
Econometrics, 2, 422-450.
J.Dhaene & M. Goovaerts (1996). Dependency of risks and stop-loss order.ASTIN Bulletin, 26,201-212. D. Drouet-Mari & S. Kotz (2001). Correlation and Dependence. Imperial College Press, London. P. Embrechts, A. McNeil & D. Straumann (2000). Correlation and dependency in risk management: Prop erties and pitfalls. Risk Management: Value at Risk and Beyond (M.A. H. Dempster, ed.), Cambridge
University J.-D. Fermanian, Bernoulli, J.-D. Fermanian 5, 25-54. Press. D. Radulovic 10,847-860. & O. Scaillet (2003). Nonparametric estimation of copulas for time series. Journal of Risk, & M. J.Wegkamp (2004). Weak convergence of empirical copula processes.
E. W. Frees & E. A. Valdez (1998). Understanding relationships using copulae. North American Actuarial
Journal,!, 1-25.
P. G?nssler & W. Stute (1987). Seminaron Empirical Processes (Schlo?Mickein, D?sseldorf 8-13,1985). DMV Seminar 9, Birkh?user, Basel. C. Genest (1987). Frank's family of bivariate distributions. Biometrika, 1A, 549-555.
September
P. Georges, A.-G. Lamy, E. Nicolas, G. Quibel & T. Roncalli (2001). Multivariate Survival Modelling: A Unified Approach with Copulas. GRO Working Paper, Cr?dit Lyonnais, Paris, France. D. Glidden (1999). Checking the adequacy of the gamma frailty model for multivariate failure times.
Biometrika, 86, 381-393. (2003). Indirect inference, Statistics, nuisance 21,122-132. is not identified under the null hypothesis. Econo parameter and threshold moving average models. A. Guay & O. Scaillet Journal of Business B. Hansen metrica, (1996).
Inference
a nuisance
parameter
64,413-430.
S. Klugman & R. Parsa (1999). Fitting bivariate loss distributions with copulas. Insurance: Mathematics
and Economics, 24,139-148.
H. Joe (1997). Multivariate Models and Dependence Concepts. Chapman & Hall, London. C. D. Lai & M. Xie (2003). Concepts of stochastic dependence in reliability analysis. InHandbook of Reliability Engineering (H. Pham, ed.), Springer, New York, pp. 141-156.
E. L. Lehmann (1966). Some concepts of dependence. The Annals of Mathematical Statistics, 37, 1137?
1153. E. L. Lehmann (1986). Testing Statistical Hypotheses. Second edition. Wiley, New York.
2005
427
H. Levy (1992). Stochastic dominance and expected utility: Survey and analysis. Management Science, 38, 555-593.
O. Linton, E. Maasoumi schemes. Review & Y.-J. Whang Consistent (2005). testing Studies, 72,735-765. of Economic for stochastic dominance under general
sampling
D. L. McFadden (1989). Testing for stochastic dominance. In Studies in theEconomics of Uncertainty (T.
Fomby & T. Seo, eds.), Springer, New York, pp. 113-134.
R. B. Nelsen (1999). An Introduction toCopulas. Lecture Notes in Statistics 139, Springer, New York. C. M. Newman (1984). Asymptotic independence and limit theorems for positively and negatively depen dent random variables. In Inequalities in Statistics and Probability (Y. L. Tong, ed.), IMS Lecture
Notes-Monographs Series, vol. 5, pp. 127-140.
Y. Nikitin (1995). Asymptotic Efficiency ofNonparametric Tests. Cambridge University Press. M. Shaked & G J. Shanthikumar (1994). Stochastic Orders and theirApplications. Academic Press, New
York.
B. Silverman (1986). Density Estimation for Statistics and Data Analysis. Chapman & Hall, London. A. Sklar (1959). Fonctions de r?partition ? n dimensions et leursmarges. Publications de l'Institut de statistique de l'Universit? de Paris, 8, 229-231. W. Stute (1984). The oscillation behavior of empirical processes: themultivariate case. The Annals of Probability, 12, 361-379. V. Tsirel'son (1975). The density of the distribution of themaximum of a Gaussian process. Theory of Probability and theirApplications, 16,847-856.
A. van der Vaart & J. A. Wellner (1996). Weak Convergence and Empirical Processes. Springer, New York.
26
May 2005
2004
Olivier SCAILLET: scaillet? hec.unige.cn HEC Gen?ve and FAME, Universit? de Gen?ve
Boulevard Carl-Vogt, 102