11

I hope this one is self-explanatory, but let me know if something is unclear: Is there a multivariate version of the Weibull distribution?

Glen_b
  • 257,508
  • 32
  • 553
  • 939
robguinness
  • 238
  • 1
  • 9
  • Apparently yes: http://onlinelibrary.wiley.com/doi/10.1002/0470011815.b2a13058/abstract I assume you did do a Google search on "multivariate Weibull". It would make it easier to help you if you told us specifically what about the Google results was not helpful to you or what you are looking for in addition. – Stephan Kolassa Nov 23 '12 at 08:08
  • Thanks. Yes, I did Google in the hope of finding an answer. I found, for example, this: http://196.1.114.11/ddh/P17.pdf but I didn't understand much of the notation. I am looking for a clear explanation of its form(s), targeted for someone with a basic introduction to statistics. – robguinness Nov 23 '12 at 09:44

1 Answers1

9

There are several in the literature.

As for what makes one suitable for your purpose, that rather depends on the purpose.

This book:

Continuous Multivariate Distributions, Models and Applications By Samuel Kotz, N. Balakrishnan, Norman L. Johnson

has some Multivariate Weibull models and is probably where I'd start.

With the use of copulas, there will be an infinite number of multivariate Weibull distributions; copulas are effectively multivariate distributions with uniform margins. You convert to or from a corresponding multivariate distribution with arbitrary continuous margins by transforming the marginals.

That way, general kinds of dependence structure can be accommodated.

Glen_b
  • 257,508
  • 32
  • 553
  • 939
  • 2
    Could you point me to some good references? – robguinness Nov 23 '12 at 09:40
  • 1
    This is potentially a good answer. Would you like to elaborate a bit your comment? Particularly on the use of copulas. Otherwise this belongs to the *comments* section. –  Nov 23 '12 at 11:14
  • @Procrastinator Fair enough; I have extended it to something more like an answer. – Glen_b Nov 23 '12 at 14:20