Extreme value statistics Problems of extrapolating to values we have no data about Question: Question: Can this be done at all? unusually large or small.

Slides:



Advertisements
Similar presentations
Introduction to modelling extremes
Advertisements

Introduction to modelling extremes Marian Scott (with thanks to Clive Anderson, Trevor Hoey) NERC August 2009.
Hydrologic Statistics Reading: Chapter 11, Sections 12-1 and 12-2 of Applied Hydrology 04/04/2006.
Hydrologic Statistics
Probability distribution functions Normal distribution Lognormal distribution Mean, median and mode Tails Extreme value distributions.
Extremes ● An extreme value is an unusually large – or small – magnitude. ● Extreme value analysis (EVA) has as objective to quantify the stochastic behavior.
Evaluating Hypotheses
WFM 5201: Data Management and Statistical Analysis
Intro to Statistics for the Behavioral Sciences PSYC 1900
Slide Slide 1 Copyright © 2007 Pearson Education, Inc Publishing as Pearson Addison-Wesley. Created by Tom Wegleitner, Centreville, Virginia Section 5-2.
Some standard univariate probability distributions
Some standard univariate probability distributions
Lecture 7 1 Statistics Statistics: 1. Model 2. Estimation 3. Hypothesis test.
Some standard univariate probability distributions
Conditional Distributions and the Bivariate Normal Distribution James H. Steiger.
1 10. Joint Moments and Joint Characteristic Functions Following section 6, in this section we shall introduce various parameters to compactly represent.
Flood Frequency Analysis
1 Chapter 12 Introduction to Statistics A random variable is one in which the exact behavior cannot be predicted, but which may be described in terms of.
Copyright © 2013, 2009, and 2007, Pearson Education, Inc. Chapter 12 Analyzing the Association Between Quantitative Variables: Regression Analysis Section.
Review of Statistical Inference Prepared by Vera Tabakova, East Carolina University ECON 4550 Econometrics Memorial University of Newfoundland.
5-2 Probability Distributions This section introduces the important concept of a probability distribution, which gives the probability for each value of.
Probability distribution functions
1 Statistical Analysis - Graphical Techniques Dr. Jerrell T. Stracener, SAE Fellow Leadership in Engineering EMIS 7370/5370 STAT 5340 : PROBABILITY AND.
Scaling forms for Relaxation Times of the Fiber Bundle Model S. S. Manna In collaboration with C. Roy,S. Kundu and S. Pradhan Satyendra Nath Bose National.
Some standard univariate probability distributions Characteristic function, moment generating function, cumulant generating functions Discrete distribution.
Traffic Modeling.
February 3, 2010 Extreme offshore wave statistics in the North Sea.
1 Institute of Engineering Mechanics Leopold-Franzens University Innsbruck, Austria, EU H.J. Pradlwarter and G.I. Schuëller Confidence.
R. Kass/W03P416/Lecture 7 1 Lecture 7 Some Advanced Topics using Propagation of Errors and Least Squares Fitting Error on the mean (review from Lecture.
Success depends upon the ability to measure performance. Rule #1:A process is only as good as the ability to reliably measure.
Going to Extremes: A parametric study on Peak-Over-Threshold and other methods Wiebke Langreder Jørgen Højstrup Suzlon Energy A/S.
Extrapolation of Extreme Response for Wind Turbines based on Field Measurements Authors: Henrik Stensgaard Toft, Aalborg University, Denmark John Dalsgaard.
Lab 3b: Distribution of the mean
Regression Regression relationship = trend + scatter
ELEC 303 – Random Signals Lecture 18 – Classical Statistical Inference, Dr. Farinaz Koushanfar ECE Dept., Rice University Nov 4, 2010.
INTRODUCTORY STUDY : WATER INDICATORS AND STATISTICAL ANALYSIS OF THE HYDROLOGICAL DATA EAST OF GUADIANA RIVER by Nikolas Kotsovinos,P. Angelidis, V. Hrissanthou,
Exploratory Tools for Spatial Data: Diagnosing Spatial Autocorrelation Main Message when modeling & analyzing spatial data: SPACE MATTERS! Relationships.
Scaling functions for finite-size corrections in EVS Zoltán Rácz Institute for Theoretical Physics Eötvös University Homepage:
Extreme value statistics Problems of extrapolating to values we have no data about Question: Question: Can this be done at all? unusually large or small.
Computer Vision Lecture 6. Probabilistic Methods in Segmentation.
Scaling functions for finite-size corrections in EVS Zoltán Rácz Institute for Theoretical Physics Eötvös University Homepage:
Extreme Value Prediction in Sloshing Response Analysis
Stracener_EMIS 7305/5305_Spr08_ Reliability Data Analysis and Model Selection Dr. Jerrell T. Stracener, SAE Fellow Leadership in Engineering.
Université d’Ottawa / University of Ottawa 2001 Bio 8100s Applied Multivariate Biostatistics L1a.1 Lecture 1a: Some basic statistical concepts l The use.
Copyright © 2010, 2007, 2004 Pearson Education, Inc. All Rights Reserved. Section 5-1 Review and Preview.
University of Colorado Boulder ASEN 5070: Statistical Orbit Determination I Fall 2015 Professor Brandon A. Jones Lecture 14: Probability and Statistics.
Probability distributions
Chris Ferro Climate Analysis Group Department of Meteorology University of Reading Extremes in a Varied Climate 1.Significance of distributional changes.
Spatial Point Processes Eric Feigelson Institut d’Astrophysique April 2014.
Week 21 Order Statistics The order statistics of a set of random variables X 1, X 2,…, X n are the same random variables arranged in increasing order.
In Bayesian theory, a test statistics can be defined by taking the ratio of the Bayes factors for the two hypotheses: The ratio measures the probability.
1 Statistical Analysis - Graphical Techniques Dr. Jerrell T. Stracener, SAE Fellow Leadership in Engineering EMIS 7370/5370 STAT 5340 : PROBABILITY AND.
Week 21 Statistical Assumptions for SLR  Recall, the simple linear regression model is Y i = β 0 + β 1 X i + ε i where i = 1, …, n.  The assumptions.
1 Ka-fu Wong University of Hong Kong A Brief Review of Probability, Statistics, and Regression for Forecasting.
Application of Extreme Value Theory (EVT) in River Morphology
Modeling and Simulation CS 313
Probability plots.
Inference for Regression
Simple Linear Regression
Extreme Value Prediction in Sloshing Response Analysis
Modeling and Simulation CS 313
Parameter Estimation 主講人:虞台文.
Statistical Inference
Flood Frequency Analysis
Multi-dimensional likelihood
Simple Linear Regression - Introduction
REGRESSION.
Hydrologic Statistics
Learning Theory Reza Shadmehr
Parametric Methods Berlin Chen, 2005 References:
Presentation transcript:

Extreme value statistics Problems of extrapolating to values we have no data about Question: Question: Can this be done at all? unusually large or small ~100 years (data) ~500 years (design) winds How long will it stand?

Extreme value paradigm is measured: Question: Question: What is the distribution of the largest number? Logics: Assume something about Use limit argument: E.g. independent, identically distributed Family of limit distributions (models) is obtained Calibrate the family of models by the measured values of parent distribution

An example of extreme value statistics The 1841 sea level benchmark (centre) on the `Isle of the Dead', Tasmania. According to Antarctic explorer, Capt. Sir James Clark Ross, it marked mean sea level in Data plots here and below are from Stuart Coles: An Introduction to Statistical Modeling of Extreme Values Recurrence time: If then the maximum will exceed in T years.

F F 1.5cm 63 fibers The weakest link problem F

Problem of trends I Variables may be non-identically distributed. Sea level seems to grow.

Problem of trends II Athletes run now faster than 30 years ago.

Problem of correlations I Maximum sea level depends, or at least is correlated to other variables.

Problem of correlations II Multivariate extremes

Problem of second-, third-, …, largest values

Problem of exceeding a threshold

Problem of deterministic background processes

Problem of the right choice of variables

Problem of spatial correlations

is measured: Fisher-Tippett-Gumbel distribution I Assumption: Independent, identically distributed random variables with parent distribution 1 st question: 1 st question: Can we estimate ? Note: 2 nd question: 2 nd question: Can we estimate ? Homework: Carry out the above estimates for a Gaussian parent distribution !

is measured: Fisher-Tippett-Gumbel distribution II Assumption: Independent, identically distributed random variables with parent distribution Question: Question: Can we calculate ? Probability of : Expected that this result does not depend on small details of. FTG density function

Fisher-Tippett-Gumbel distribution III Question: Question: What is the „fitting to FTG” procedure? We do not know the parent distribution! is measured. The shift in is not known! The scale of can be chosen at will. Fitting to: Asymptotes: -1 largest smallest Important: Important: In the simplest EVS paradigm only linear change of variables is allowed. Without this restriction any distribution could be obtained!

FTG function and fitting

FTG function and fitting: Logscale See example on fitting.

Fisher-Tippett-Fréchet distribution I Parent distribution: Power decay is measured. 1 st question: 1 st question: Can we estimate the typical maximum? 2 nd question: 2 nd question: Can we estimate the deviation? If it exists! The maximum is on the same scale as the deviation.

Fisher-Tippett-Fréchet distribution II Question: Question: Can we calculate ? Probability of : FTF density function is measured: Assumption: Independent, identically distributed random variables with parent For large :

Fisher-Tippett-Fréchet distribution III The origin and the scale of x can be chosen at will: The function to fit for x>a is Note that forthere is no average! The kth moment does not exist for in is not known!

FTF density function for

is measured: Finite cutoff: Weibull distribution I Assumption: Independent, identically distributed random variables with parent distribution 1 st question: 1 st question: Can we estimate ? 2 nd question: 2 nd question: Can we estimate ?

Weibull distribution II parent distribution Question: Question: Can we calculate ? Probability of : Weibull density function is measured: Assumption: Independent, identically distributed random variables with if

Weibull distribution III parent distribution is measured. are not known! in is not known! Fitting to and possibly The scaleofcan be chosen at will.

Weibull function and fitting

Notes about the Tmax homework Introduce scaled variables common to all data sets Find Average and width of distribution so all data can be analyzed together. ? ? ? ? ? ? ? ? What kind of conclusions can be drawn?