Correlation measures: Robustness and performance in simulations

Srinivas Sriramula, Devdas Menon, A. Meher Prasad

Research output: Contribution to conferenceUnpublished paperpeer-review


Many civil engineering studies involve uncertainties arising at various levels and the rational modeling of such uncertainties requires effective simulation methods involving accurate dependencies. In this paper, the potential measures of dependence are described and their robustness is considered using the maximum bias estimates. The copula based multivariate distribution models are used in studying the bias at varying levels of data contamination with a point mass distribution. Two examples are considered pertinent to civil engineering applications and the database is considered to represent the classes of elliptical and non-elliptical marginal distribution combinations. The dependence measures are found to be biased depending on the marginal distribution type and the associated data scatter highlighting the need for appropriate dependence modelling in simulations. It is required to popularise the use of copula functions in engineering applications to consider the variable uncertainties effectively with the appropriate correlations.
Original languageEnglish
Number of pages6
Publication statusPublished - 2007
Event10th Int. Conf. on Applications of Statistics and Probability in Civil Engineering (ICASP10) - Tokyo, Japan
Duration: 31 Jul 200731 Jul 2007


Conference10th Int. Conf. on Applications of Statistics and Probability in Civil Engineering (ICASP10)


  • simulations
  • product moment correlation
  • rank correlation
  • maximum bias curves
  • copula functions


Dive into the research topics of 'Correlation measures: Robustness and performance in simulations'. Together they form a unique fingerprint.

Cite this