Introduction to Extreme Value Analysis in R – Part 1: Software

Hey there!

Today I’d like to start a series of some posts concerning extreme value analysis using R.

Basically, there are several very useful packages in R which provide methods and functions for extreme value analysis. Information on different software (including all relevant R packages) for extreme value analysis can of course be found at the R Task View on Extreme Value Analysis as well as on Eric Gilleland’s website. In addition, Gilleland, Ribatet & Stephenson have published A software review for extreme value analysis back in 2012, which provides a comprehensive overview of the most important software tools related to this topic.

Even though there are several R packages that provide nice solutions for different estimation methods (e.g. maximum likelihood fitting, L-moments estimation, Bayesian estimation methods), different approaches (i.e. block maxima method or peak over threshold method) or different tasks (e.g. calculation of return levels/periods, simulation, use of copulas, assessment of non-stationary time series, multivariate analyses), I reckon that the extRemes package contains most of the relevant methods for performing a sound extreme value analysis. fevd{extRemes} is a very powerful (albeit imho barely readable) function that allows us to apply the most common methods for distribution fitting and parameter estimation.

Apart from the extRemes package, I can recommend the package fExtremes, which is developed by Rmetrics and mainly aims at the analysis of extreme financial market data. It contains some well-written, useful functions.

In addition, the ismev package, which is based on Stuart Coles’ book An Introduction to Statistical Modeling of Extreme Values, provides many basic methods for performing extreme value analyses and is especially useful for delving into this subject in combination with Coles’ book.

I will prepare some examples covering both block maxima and peak over threshold approach as well as the most important estimation methods.

Be sure to stay tuned!

Cheers,
Matthias

About This Author

Matthias studied Environmental Information Management at the University of Natural Resources and Life Sciences Vienna and holds a PhD in environmental statistics. The focus of his thesis was on the statistical modelling of rare (extreme) events as a basis for vulnerability assessment of critical infrastructure. He is working at the Austrian national weather and geophysical service (ZAMG) and at the Institute of Mountain Risk Engineering at BOKU University. He currently focuses the (statistical) assessment of adverse weather events and natural hazards, and disaster risk reduction. His main interests are statistical modelling of environmental phenomena as well as open source tools for data science, geoinformation and remote sensing.

3 Comments

You can post comments in this post.

  • Very interesting topic! I am quite curious and very much looking forward to this new segment!

    Martin 8 years ago Reply

  • What about library evir? Aslo lookin forward!

    Gokhan 8 years ago Reply

    • As a matter of fact there are several useful R packages featuring tools for extreme value analysis. Even though packages like evir, evd, fExtremes, lmom, POT, ismev, etc. are nice packages, extRemes is my package of choice in most cases. While its clear and simple structrue also allows beginners to get in touch with extreme value analysis (which is in fact the intention of my posts), it is also a very comprehensive package as far as the implemented methods are concerned. Thus, it is also a good tool for advanced users. Nevertheless, other packages may provide better options for certain tasks, which is why I switched to using fExtremes lately.

      Matthias 8 years ago Reply

Post A Reply

CAPTCHA

*