Characterizing Non- Gaussianities or How to tell a Dog from an Elephant Jesús Pando DePaul University.

Slides:



Advertisements
Similar presentations
ELEN 5346/4304 DSP and Filter Design Fall Lecture 15: Stochastic processes Instructor: Dr. Gleb V. Tcheslavski Contact:
Advertisements

Probability Distributions CSLU 2850.Lo1 Spring 2008 Cameron McInally Fordham University May contain work from the Creative Commons.
Use of moment generating functions. Definition Let X denote a random variable with probability density function f(x) if continuous (probability mass function.
Pierfrancesco Cacciola Senior Lecturer in Civil Engineering ( Structural Design ) School of Environment and Technology, University of Brighton, Cockcroft.
AP Statistics Chapter 16. Discrete Random Variables A discrete random variable X has a countable number of possible values. The probability distribution.
Review of Basic Probability and Statistics
Statistics & Modeling By Yan Gao. Terms of measured data Terms used in describing data –For example: “mean of a dataset” –An objectively measurable quantity.
Quality Control Procedures put into place to monitor the performance of a laboratory test with regard to accuracy and precision.
Descriptive statistics Experiment  Data  Sample Statistics Experiment  Data  Sample Statistics Sample mean Sample mean Sample variance Sample variance.
Environmentally Conscious Design & Manufacturing (ME592) Date: May 5, 2000 Slide:1 Environmentally Conscious Design & Manufacturing Class 25: Probability.
Probability and Statistics Review
The moment generating function of random variable X is given by Moment generating function.
Continuous Random Variables and Probability Distributions
Probability Distributions and Frequentist Statistics “A single death is a tragedy, a million deaths is a statistic” Joseph Stalin.
1 ECE310 – Lecture 23 Random Signal Analysis 04/27/01.
Introduction To Signal Processing & Data Analysis
(1) A probability model respecting those covariance observations: Gaussian Maximum entropy probability distribution for a given covariance observation.
Lecture II-2: Probability Review
Probability Distributions W&W Chapter 4. Continuous Distributions Many variables we wish to study in Political Science are continuous, rather than discrete.
1 10. Joint Moments and Joint Characteristic Functions Following section 6, in this section we shall introduce various parameters to compactly represent.
Normal and Sampling Distributions A normal distribution is uniquely determined by its mean, , and variance,  2 The random variable Z = (X-  /  is.
NIPRL Chapter 2. Random Variables 2.1 Discrete Random Variables 2.2 Continuous Random Variables 2.3 The Expectation of a Random Variable 2.4 The Variance.
Review of Probability.
Prof. SankarReview of Random Process1 Probability Sample Space (S) –Collection of all possible outcomes of a random experiment Sample Point –Each outcome.
Moment Generating Functions 1/33. Contents Review of Continuous Distribution Functions 2/33.
Introduction l Example: Suppose we measure the current (I) and resistance (R) of a resistor. u Ohm's law relates V and I: V = IR u If we know the uncertainties.
Error Analysis Accuracy Closeness to the true value Measurement Accuracy – determines the closeness of the measured value to the true value Instrument.
Overview Summarizing Data – Central Tendency - revisited Summarizing Data – Central Tendency - revisited –Mean, Median, Mode Deviation scores Deviation.
Moment Generating Functions
Chapter Twelve Census: Population canvass - not really a “sample” Asking the entire population Budget Available: A valid factor – how much can we.
1 LES of Turbulent Flows: Lecture 1 Supplement (ME EN ) Prof. Rob Stoll Department of Mechanical Engineering University of Utah Fall 2014.
1 Lesson 3: Choosing from distributions Theory: LLN and Central Limit Theorem Theory: LLN and Central Limit Theorem Choosing from distributions Choosing.
Functions of Random Variables. Methods for determining the distribution of functions of Random Variables 1.Distribution function method 2.Moment generating.
K. Shum Lecture 16 Description of random variables: pdf, cdf. Expectation. Variance.
Continuous Distributions The Uniform distribution from a to b.
Chapter 7 Sampling and Sampling Distributions ©. Simple Random Sample simple random sample Suppose that we want to select a sample of n objects from a.
Biostatistics Unit 5 – Samples. Sampling distributions Sampling distributions are important in the understanding of statistical inference. Probability.
LES of Turbulent Flows: Lecture 2 (ME EN )
One Random Variable Random Process.
CCN COMPLEX COMPUTING NETWORKS1 This research has been supported in part by European Commission FP6 IYTE-Wireless Project (Contract No: )
Stats Probability Theory Summary. The sample Space, S The sample space, S, for a random phenomena is the set of all possible outcomes.
Random Variables (1) A random variable (also known as a stochastic variable), x, is a quantity such as strength, size, or weight, that depends upon a.
Expectation. Let X denote a discrete random variable with probability function p(x) (probability density function f(x) if X is continuous) then the expected.
Wavelet Spectral Analysis Ken Nowak 7 December 2010.
Probing cosmic structure formation in the wavelet representation Li-Zhi Fang University of Arizona IPAM, November 10, 2004.
Review of Probability. Important Topics 1 Random Variables and Probability Distributions 2 Expected Values, Mean, and Variance 3 Two Random Variables.
Sampling and estimation Petter Mostad
Fourier series, Discrete Time Fourier Transform and Characteristic functions.
AP Statistics Chapter 16. Discrete Random Variables A discrete random variable X has a countable number of possible values. The probability distribution.
CHAPTER 13 DETERMINING THE SIZE OF A SAMPLE. Important Topics of This Chapter Different Methods of Determining Sample size. Standard Normal Distribution.
Random Variables. Numerical Outcomes Consider associating a numerical value with each sample point in a sample space. (1,1) (1,2) (1,3) (1,4) (1,5) (1,6)
Discrete-time Random Signals
Review of Probability Concepts Prepared by Vera Tabakova, East Carolina University.
ECE-7000: Nonlinear Dynamical Systems 2. Linear tools and general considerations 2.1 Stationarity and sampling - In principle, the more a scientific measurement.
Lecture 8: Measurement Errors 1. Objectives List some sources of measurement errors. Classify measurement errors into systematic and random errors. Study.
Pattern Recognition Mathematic Review Hamid R. Rabiee Jafar Muhammadi Ali Jalali.
Geology 6600/7600 Signal Analysis 04 Sep 2014 © A.R. Lowry 2015 Last time: Signal Analysis is a set of tools used to extract information from sequences.
STATISTICS People sometimes use statistics to describe the results of an experiment or an investigation. This process is referred to as data analysis or.
Environmental Data Analysis with MatLab 2 nd Edition Lecture 22: Linear Approximations and Non Linear Least Squares.
Statistics and probability Dr. Khaled Ismael Almghari Phone No:
Sampling and Sampling Distributions
Introduction to Probability - III John Rundle Econophysics PHYS 250
k is the frequency index
Introductory Statistics and Data Analysis
Outline Introduction Signal, random variable, random process and spectra Analog modulation Analog to digital conversion Digital transmission through baseband.
Analyzing Redistribution Matrix with Wavelet
Review of Probability Concepts
MEGN 537 – Probabilistic Biomechanics Ch.3 – Quantifying Uncertainty
k is the frequency index
Chapter 2. Random Variables
Presentation transcript:

Characterizing Non- Gaussianities or How to tell a Dog from an Elephant Jesús Pando DePaul University

Contents Gaussian Signals Distinguishing among Gaussian signals The non-Gaussian domain and the failure of spectral methods Wavelets Distinguishing dogs from elephants Conclusion

Gaussian Signals Normal distributions are common, because the central limit theorem states that the sum of independent random variables with finite variance will result in normal distributions. Normal distributions are common, because the central limit theorem states that the sum of independent random variables with finite variance will result in normal distributions.

The probability density is centered at the mean and 68% of the distribution lies within the square root of the second central moment, or the standard deviation The probability density is centered at the mean and 68% of the distribution lies within the square root of the second central moment, or the standard deviation. However, distributions caused by different physical processes with different time scales can have similar means and variances.

Two Gaussian signals with mean = 0 and variance = 1.5

Spectral Methods  We need the variance as a function of scale in order to distinguish these distributions of the same mean and variance.

The Power Spectrum reveals differences

Spectral Methods Spectral techniques are effective in untangling Gaussian signals. Spectral techniques are effective in untangling Gaussian signals. The power spectrum, or variance as a function of scale, breaks down contributions of different physical processes to a signal. The power spectrum, or variance as a function of scale, breaks down contributions of different physical processes to a signal.

Non-Gaussian Signals  Non-Gaussian distributions  Non-Linear dynamics and chaos  Multi-component systems  Non-Gaussian signals are much harder to characterize and detect.  One easy way to distinguish between Gaussian and non-Gaussian signals is by the use of cumulants.

Moments and Cumulants  The n th moment of distribution having probability density f and mean µ is:  x n f(x) dx  x n f(x) dx  (x - µ) n f(x) dx is the n th central moment.  (x - µ) n f(x) dx is the n th central moment.  Cumulants are defined via moments. For the the 1st and 2nd central moment (mean and variance), cumulants are equal to moments. However, higher order cumulants are given by the recursion formula:

For instance, the first 3 cumulants defined in terms of the central moments are:For instance, the first 3 cumulants defined in terms of the central moments are:  2 =  2  3 =  3  4 =   2 2  4 =   2 2 For a normal distribution For a normal distribution  1 =  1 (mean)  1 =  1 (mean)  2 =  2 (variance)  2 =  2 (variance)  n = 0 for n > 2  n = 0 for n > 2 Cumulants are a way to detect non-Gaussian signals.Cumulants are a way to detect non-Gaussian signals. However, we are faced with same problem as before; that is, it may be possible to have two very different signals with the n th cumulant equal.However, we are faced with same problem as before; that is, it may be possible to have two very different signals with the n th cumulant equal.

Wavelet Transform  The wavelet transform is an integral transform whose basis functions are well localized in time and frequency (or space and scale).  Wavelets have become increasingly important because of the ability to localize a signal efficiently in both time and frequency.

Unlike the Fourier transform, there is no unique wavelet basis. Instead, the wavelets are defined by a function, , that is rescaled and translated:

Wavelet Properties: Most useful wavelets have compact support. Wavelets can be classified as continuous or discrete. Wavelets can be classified as continuous or discrete. The discrete wavelet transform (DWT) produces two sets of coefficients; the scaling coefficients which give a local average, and the wavelet coefficients which give the fluctuation from the local average. The discrete wavelet transform (DWT) produces two sets of coefficients; the scaling coefficients which give a local average, and the wavelet coefficients which give the fluctuation from the local average. The most useful DWT are also orthogonal. Especially, the wavelet coefficients are orthogonal in both space and scale (time and frequency). The most useful DWT are also orthogonal. Especially, the wavelet coefficients are orthogonal in both space and scale (time and frequency).

WFC’s,  j,l, measure changes from local mean. A large WFC indicates a large local fluctuation. WFC’s look suspiciously like a central moment. WFC’s look suspiciously like a central moment. As with the Fourier spectrum, we can define the Wavelet Variance Spectrum: P j = where the average is done over position l at scale j. As with the Fourier spectrum, we can define the Wavelet Variance Spectrum: P j = where the average is done over position l at scale j. With wavelets, higher order cumulant spectra are readily defined. With wavelets, higher order cumulant spectra are readily defined.

We define the third and fourth order cumulants as: S j = M j 3 /(M j 2 ) 3/2 K j = M j 4 /(M j 2 ) where M n = Thus, the wavelet gives a simple way to characterize some non-Gaussian distributions.

Gaussian distribution with power spectrum whereis constant Gaussian distribution with power spectrum, P(k) = k/(1 + a k 4 ) where a is constant.

Non-Gaussian Simulations  Clumps or valleys with a signal/noise = 2.0 and random width between 1-5 bins are embedded in a Gaussian background.  Distributions with16, 32, and 48 clumps (or valleys) are generated.  100 realizations of each is done and 95% confidence levels computed.

Cumulant,  3 Cumulant,  4

Scale-Scale Correlations The DWT cumulant spectra give a way to characterize different non-Gaussian signals. DWT measure can also give clues to the dynamics behind the non-Gaussian distributions. In scale dependent processes, one such measure is the scale-scale correlation.

The Gaussian Block model results in a final distribution that is Gaussian since it is formed at each level by Gaussian random variables. The Gaussian Block model results in a final distribution that is Gaussian since it is formed at each level by Gaussian random variables. The Branching block model results in a final distribution that is not Gaussian since each level(scale) has a memory of how it got there. The Branching block model results in a final distribution that is not Gaussian since each level(scale) has a memory of how it got there.

The usual statistical measures fail to distinguish these distributions. The usual statistical measures fail to distinguish these distributions. We introduce the scale-scale DWT correlation: We introduce the scale-scale DWT correlation: For a Gaussian distribution, C j p,p = 1 for p  2. This statistical measure can therefore detect some types of dynamics (hierarchical).

Conclusion One point measures detect non-Gaussianities, but provide limited information about the signal. One point measures detect non-Gaussianities, but provide limited information about the signal. Traditional Fourier spectral methods are not ideal for higher order cumulants. Traditional Fourier spectral methods are not ideal for higher order cumulants. Wavelets allow one to construct cumulant spectra. Wavelets allow one to construct cumulant spectra. Wavelet versatility allows for the construction of customized measures and sometimes help us to say more than just a dog is not an elephant. Wavelet versatility allows for the construction of customized measures and sometimes help us to say more than just a dog is not an elephant.