Geodesic
On the connection between cherry-tree copulas and truncated R-vine copulas
Kybernetika, Tome 53 (2017) no. 3, pp. 437-460
Cet article a éte moissonné depuis la source Czech Digital Mathematics Library

Voir la notice de l'article

Vine copulas are a flexible way for modeling dependences using only pair-copulas as building blocks. However if the number of variables grows the problem gets fastly intractable. For dealing with this problem Brechmann at al. proposed the truncated R-vine copulas. The truncated R-vine copula has the very useful property that it can be constructed by using only pair-copulas and a lower number of conditional pair-copulas. In our earlier papers we introduced the concept of cherry-tree copulas. In this paper we characterize the relation between cherry-tree copulas and truncated R-vine copulas. It turns out that the concept of cherry-tree copula is more general than the concept of truncated R-vine copula. Although both contain in their expressions conditional independences between the variables, the truncated R-vines constructed in greedy way do not exploit the existing conditional independences in the data. We give a necessary and sufficient condition for a cherry-tree copula to be a truncated R-vine copula. We introduce a new method for truncated R-vine modeling. The new idea is that in the first step we construct the top tree by exploiting conditional independences for finding a good-fitting cherry-tree of order $k$. If this top tree is a tree in an R-vine structure then this will define a truncated R-vine at level $k$ and in the second step we construct a sequence of trees which leads to it. If this top tree is not a tree in an R-vine structure then we can transform it into such a tree at level $k+1$ and then we can again apply the second step. The second step is performed by a backward construction named Backward Algorithm. This way the cherry-tree copulas always can be expressed by pair-copulas and conditional pair-copulas.
Vine copulas are a flexible way for modeling dependences using only pair-copulas as building blocks. However if the number of variables grows the problem gets fastly intractable. For dealing with this problem Brechmann at al. proposed the truncated R-vine copulas. The truncated R-vine copula has the very useful property that it can be constructed by using only pair-copulas and a lower number of conditional pair-copulas. In our earlier papers we introduced the concept of cherry-tree copulas. In this paper we characterize the relation between cherry-tree copulas and truncated R-vine copulas. It turns out that the concept of cherry-tree copula is more general than the concept of truncated R-vine copula. Although both contain in their expressions conditional independences between the variables, the truncated R-vines constructed in greedy way do not exploit the existing conditional independences in the data. We give a necessary and sufficient condition for a cherry-tree copula to be a truncated R-vine copula. We introduce a new method for truncated R-vine modeling. The new idea is that in the first step we construct the top tree by exploiting conditional independences for finding a good-fitting cherry-tree of order $k$. If this top tree is a tree in an R-vine structure then this will define a truncated R-vine at level $k$ and in the second step we construct a sequence of trees which leads to it. If this top tree is not a tree in an R-vine structure then we can transform it into such a tree at level $k+1$ and then we can again apply the second step. The second step is performed by a backward construction named Backward Algorithm. This way the cherry-tree copulas always can be expressed by pair-copulas and conditional pair-copulas.
DOI : 10.14736/kyb-2017-3-0437
Classification : 60C05, 62H05
Keywords: copula; conditional independences; Regular-vine; truncated vine; cherry-tree copula 
@article{10_14736_kyb_2017_3_0437,
     author = {Kov\'acs, Edith and Sz\'antai, Tam\'as},
     title = {On the connection between cherry-tree copulas and truncated {R-vine} copulas},
     journal = {Kybernetika},
     pages = {437--460},
     year = {2017},
     volume = {53},
     number = {3},
     doi = {10.14736/kyb-2017-3-0437},
     mrnumber = {3684679},
     zbl = {06819617},
     language = {en},
     url = {http://geodesic.mathdoc.fr/articles/10.14736/kyb-2017-3-0437/}
}
TY  - JOUR
AU  - Kovács, Edith
AU  - Szántai, Tamás
TI  - On the connection between cherry-tree copulas and truncated R-vine copulas
JO  - Kybernetika
PY  - 2017
SP  - 437
EP  - 460
VL  - 53
IS  - 3
UR  - http://geodesic.mathdoc.fr/articles/10.14736/kyb-2017-3-0437/
DO  - 10.14736/kyb-2017-3-0437
LA  - en
ID  - 10_14736_kyb_2017_3_0437
ER  - 
%0 Journal Article
%A Kovács, Edith
%A Szántai, Tamás
%T On the connection between cherry-tree copulas and truncated R-vine copulas
%J Kybernetika
%D 2017
%P 437-460
%V 53
%N 3
%U http://geodesic.mathdoc.fr/articles/10.14736/kyb-2017-3-0437/
%R 10.14736/kyb-2017-3-0437
%G en
%F 10_14736_kyb_2017_3_0437
Kovács, Edith; Szántai, Tamás. On the connection between cherry-tree copulas and truncated R-vine copulas. Kybernetika, Tome 53 (2017) no. 3, pp. 437-460. doi: 10.14736/kyb-2017-3-0437

[1] Aas, K., Czado, C., Frigessi, A., Bakken, H.: Pair-copula constructions of multiple dependence. Insur. Math. Econom. 44 (2009), 182-198. | DOI | MR

[2] Acar, E. F., Genest, C., Nešlehová, J.: Beyond simplified pair-copula constructions. J. Multivariate Anal. 110 (2012), 74-90. | DOI | MR

[3] Bauer, A., Czado, C., Klein, T.: Pair-copula construction for non-Gaussian DAG models. Canad. J. Stat. 40 (2012), 1, 86-109. | DOI | MR

[4] Bedford, T., Cooke, R.: Probability density decomposition for conditionally dependent random variables modeled by vines. Ann. Math. Artif. Intell. 32 (2001), 245-268. | DOI | MR

[5] Bedford, T., Cooke, R.: Vines - a new graphical model for dependent random variables. Ann. Statist. 30 (2002), 4, 1031-1068. | DOI | MR

[6] Brechmann, E. C., Czado, C., Aas, K.: Truncated regular vines in high dimensions with applications to financial data. Canad. J. Statist. 40 (2012), 1, 68-85. | DOI | MR

[7] Bukszár, J., Prékopa, A.: Probability bounds with cherry trees. Math. Oper. Res. 26 (2001), 174-192. | DOI | MR

[8] Bukszár, J., Szántai, T.: Probability bounds given by hypercherry trees. Optim. Methods Software 17 (2002), 409-422. | DOI | MR

[9] Cover, T. M., Thomas, J. A.: Elements of Information Theory. Wiley Interscience, New York 1991. | DOI | MR

[10] Czado, C.: Pair-copula constructions of multivariate copulas. In: Copula Theory and Its Applications (P. Jaworski, F. Durante, W. Härdle, and T. Rychlik, eds.), Springer, Berlin 2010. | DOI | MR

[11] Dissman, J., Brechmann, E. C., Czado, C., Kurowicka, D.: Selecting and estimating regular vine copulae and application to financial returns. Comput. Statist. Data Anal. 59 (2013), 52-69. | DOI | MR

[12] Hanea, A., Kurowicka, D., Cooke, R.: Hybrid method for quantifying and analyzing Bayesian belief networks. Qual. Reliab. Engrg. 22 (2006), 708-729. | DOI

[13] Haff, I. Hobaek, Aas, K., Frigessi, A.: On the simplified pair-copula construction - simply useful or too simplistic?. J. Multivariate Anal. 101 (2010), 5, 1296-1310. | DOI | MR

[14] Haff, I. Hobaek, Segers, J.: Nonparametric estimation of pair-copula constructions with the empirical pair-copula. 2010. | arXiv

[15] Hobaek-Haff, I., Aas, K., Frigessi, A., Lacal, V.: Structure learning in Bayesian Networks using regular vines. Computat. Statist. Data Anal. 101 (2016), 186-208. | DOI | MR

[16] Joe, H.: Multivariate Models and Dependence Concepts. Chapman and Hall, London 1997. | DOI | MR | Zbl

[17] Kovács, E., Szántai, T.: On the approximation of discrete multivariate probability distribution using the new concept of $t$-cherry junction tree. Lect. Notes Economics Math. Systems 633, Proc. IFIP/IIASA/GAMM Workshop on Coping with Uncertainty, Robust Solutions, 2008, IIASA, Laxenburg 2010, pp. 39-56. | DOI | MR

[18] Kovács, E., Szántai, T.: Multivariate copula expressed by lower dimensional copulas. 2010. | arXiv

[19] Kovács, E., Szántai, T.: Hypergraphs in the characterization of regular-vine copula structures. In: Proc. 13th International Conference on Mathematics and its Applications, Timisoara 2012(a), pp. 335-344.

[20] Kovács, E., Szántai, T.: Vine copulas as a mean for the construction of high dimensional probability distribution associated to a Markov network. 2012(b). | arXiv

[21] Kurowicka, D., Cooke, R.: The vine copula method for representing high dimensional dependent distributions: Application to continuous belief nets. In: Proc. 2002 Winter Simulation Conference 2002, pp. 270-278. | DOI

[22] Kurowicka, D., Cooke, R. M.: Uncertainty Analysis with High Dimensional Dependence Modelling. John Wiley, Chichester 2006. | DOI | MR

[23] Kurowicka, D.: Optimal truncation of vines. In: Dependence-Modeling - Handbook on Vine Copulas (D. Kurowicka and H. Joe, eds.), Word Scientific Publishing, Singapore 2011. | MR

[24] Lauritzen, S. L., Spiegelhalter, D. J.: Local Computations with probabilites on graphical structures and their application to expert systems. J. Roy. Statist. Soc. B 50 (1988), 157-227. | MR

[25] Lauritzen, S. L.: Graphical Models. Clarendon Press, Oxford 1996. | MR

[26] Szántai, T., Kovács, E.: Hypergraphs as a mean of discovering the dependence structure of a discrete multivariate probability distribution. In: Proc. Conference Applied Mathematical Programming and Modelling (APMOD), Bratislava 2008, Ann. Oper. Res. 193 (2012), 1, 71-90. | DOI | MR

[27] Whittaker, J.: Graphical Models in Applied Multivariate Statistics. John Wiley and Sons, 1990. | MR

Cité par Sources :