Initialization for NORTA: Generation of random vectors with specified marginals and correlations

Authors
Citation
Hf. Chen, Initialization for NORTA: Generation of random vectors with specified marginals and correlations, INFORMS J C, 13(4), 2001, pp. 312-331
Citations number
47
Language
INGLESE
art.tipo
Article
Categorie Soggetti
Computer Science & Engineering
Journal title
INFORMS JOURNAL ON COMPUTING
ISSN journal
1091-9856 → ACNP
Volume
13
Issue
4
Year of publication
2001
Pages
312 - 331
Database
ISI
SICI code
1091-9856(200123)13:4<312:IFNGOR>2.0.ZU;2-J
Abstract
We propose a specific method for generating n-dimensional random vectors wi th given marginal distributions and correlation matrix. The method uses the NORTA (NOR-mal To Anything) approach, which generates a standard normal ra ndom vector and then transforms it into a random vector with specified marg inal distributions. During initialization, n(n - 1)/2 nonlinear equations n eed to be solved to ensure that the generated random vector has the specifi ed correlation structure. To solve these equations, we apply retrospective approximation, a generic stochastic root-finding algorithm, with slight cha nges. Internal control variates are used to estimate function values. Empir ical comparisons show that the control-variate variance-reduction technique improves the algorithm's convergence speed as well as its robustness. Simu lation results for a variety of marginal distributions and correlation matr ices are also presented.