Skip to content
All library documents

Fitting Skewed Copulas with Time-Varying Dependence

Article Quant Q&A · Author: Kondo
Read the original

Summary

The post asks how to fit a skew-normal or skew-t copula to data while allowing the correlation matrix to change over time. It highlights an implementation challenge: the author has not found a suitable approach in R or Matlab and is open to multiple definitions of skew distributions from the literature.

The only response points to the R copula package as a possible starting point and suggests consulting its documentation and a package paper. It does not give a fitting procedure, explain how to model time-varying dependence, or demonstrate that the package supports the requested skewed copulas. The post is therefore useful mainly as a pointer to software for copula work and as a reminder that skew-normal and skew-t distributions have competing definitions. Any implementation would need to verify the package’s capabilities and choose a specific distribution definition before estimating the model.

Key ideas

  • The question concerns skew-normal or skew-t copulas with time-varying correlation.
  • Different references define skew-t distributions in different ways, so model specification matters.
  • The response recommends exploring the R copula package and its documentation.
  • No estimation method or evidence of support for the requested dynamic skew copulas is provided.

Tags

StatisticsRisk managementMachine learning
Full text
# How to fit a skew normal/t copula to data?


# How to fit a skew normal/t copula to data?












I want to use either the skew normal copula or the skew t copula with a time-varying correlation matrix. But so far I haven't found any way to implement this either in R or Matlab.

Would anyone be able to help, does anyone maybe have a code available from some previous work? Or maybe do you know someone who works with copulas for risk management purposes?

In the literature I see that there are a few definitions available for these distributions (the skew t for example is defined differently in Azzalini & Capitanio (2003), Demarta, S. and A. J. McNeil (2005) or Sahu, Dey & Branco (2003)). Any definition of the skew normal or t you find in the literature is fine with me.

Any help is appreciated!

## Answer by Robert (score 0)

https://quant.stackexchange.com/a/28101

Have you look at `copula` `package`! Maybe you could get ideias from it https://www.jstatsoft.org/article/view/v021i04/v21i04.pdf

http://finzi.psych.upenn.edu/R/library/copula/html/copula-package.html

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.