The Spectral Density Estimation of Stationary Time Series with Missing Data

Size: px
Start display at page:

Download "The Spectral Density Estimation of Stationary Time Series with Missing Data"

Transcription

1 The Spectral Density Estimation of Stationary Time Series with Missing Data Jian Huang and Finbarr O Sullivan Department of Statistics University College Cork Ireland Abstract The spectral estimation of unevenly sampled data has been widely investigated in astronomical and medical areas. However the investigations are usually carried out in the context of periodicity detection and deterministic signal. Here we consider estimating the spectral density of stationary time series with missing data. An asymptotically unbiased estimation approach is proposed. The simulations are used to compare it to the classical periodogram, the Lomb periodogram (widely used for irregularly sampled data) and the SVD based periodogram. The results show that the new method substantially reduced the bias and slightly increased the variance. Overall the new approach significantly reduced the mean squared percentage error. As an example, the approach is applied to rainfall data in Irelan. Keywords: stationary time series, singular value decomposition, spectral density, the Lomb periodogram. 1 Introduction The Lomb periodogram(lomb, 1976 and Scargle, 1982) has been proposed for processing of irregularly sampled data. Its successful application to detecting periodicities in data and to estimating the spectral density of deterministic signal has been reported in literature (see, e.g. Green et al., 2002, Kaneoke and Vitek, 1996, Fortin and Mackey, 1999, and Laguna et al., 1998). Here we consider estimating the spectral density of stationary time series with missing data with focus on the estimation accuracy. This is motivated by the following facts: In climatological studies, the anomaly data can usually be modeled as stationary time series (after removing some trend and seasonal components). common in climatological data. Missing observations are 1

2 Now consider a missing data set {x t1,..., x tn } sampled from the stationary time series {X t }. The discrete Fourier transform (DFT) can be defined as NX FT(ω) = x tj exp iωt j, j=1 where t j s are non-missing sampling times and N is the number of non-missing values. The periodogram is then conventionally defined as P (ω) = = 1 2Nπ FT((ω) 2 (1) 1 N 2Nπ [( X NX x j cos(ωt j )) 2 +[( x j sin(ωt j )) 2 ]. j=1 j=1 Although the periodogram can be evaluated for any frequency, it is traditionally evaluated only at a special set of evenly spaced frequencies (the Fourier frequencies) ω k =2kπ/N k =0, 1,...,[N/2], where [N/2] is the largest integer less than or equate to N/2. It is well known that at the Fourier frequencies sinusoid components in (2) are orthogonal. However this property does not hold with missing data. To recovery the orthogonality, Lomb (1976) introduced a time delay τ, defined for each frequency ω by tan(2ωτ) = P Nj=1 sin(2ωt j ) P Nj=1 cos(2ωt j ), his redefinition of the periodogram called the Lomb periodogram is P (ω) = 1 P Nj=1 2Nπ [( x j cosω(t j τ)) 2 P Nj=1 cos 2 ω(t j τ) + (P N j=1 x j sinω(t j τ)) 2 P Nj=1 ]. sin 2 ω(t j τ) The Lomb periodogram is preferable to the classical periodogram for two reasons: it has a simple statistical behavior, and is equivalent to the reduction of the sum of squares fitting of sine waves to the data (Scargle, 1982). However the actual values are typically not changed much (Scargle, 1982). Hence we cannot expect the Lomb periodogram will lead a significant improvement in the estimation accuracy. In the non-missing case, alternatively, the classical periodogram can be obtained by solving the linear equation system resulted from fitting sine waves with the Fourier frequencies to the data. (Brockwell and Davis, 1993). Now if some values are missing, an under-determined linear equation system results (that is the number of equations less than 2

3 the number of unknown coefficients). A nature remedy is using the signgular values decomposition to solve this equation system. The SVD based approach (may called the SVD periodogram) is investigated using simulation studies. The results show that the actual values are quite close to the classical periodogram (see Section 3). In this paper we investigate a two-step spectral estimation approach. The basic idea of the method is to refine the classical estimator based on the effect on the spectral density of misssing data. This approach is described in Section 2. In Section 3 we compare the refined periodogram to the classical periodogram, the lomb periodogram and the SVD periodogram using simulated data. Conclusions and further research are given in Section 4. 2 Refined periodogram The missing data will be modeled as Y t = W t X t (2) where W t are stationary random weights with 0-1 values (0 for missing and 1 for nonmissing). {X t } and {Y t } are zero mean stationary time series with the spectral densities f X (λ) andf Y (λ), respectively. The problem is to estimate the spectral density of {X t } using observations on {Y t } and {W t }. We assume that {W t } is independent of {X t }.Thenwehave R Y (h) =EY t+h Y t = E(W t+h W t )E(X t+h X t )=(R W (h)+e 2 W )R X (h), where E 2 W and R W (h) arethemeanandtheautocovarincefunctionof{w t }, respectively. So It follows that R X (h) = R Y (h) EW 2 (1 R W (h) R W (h)+ew 2 ) f X (ω) = 1 EW 2 f Y (ω) 1 EW 2 f Y (ω) l W (ω) (3) R Where l W (ω) is the Fourier transformation of W (h). Hence we proposed to estimate R W (h)+ew 2 the spectral density of {X t } by the following two stage procedure: Step 1. Estimating the spectral density of {Y t } using the observations on {Y t }.-ˆf Y (ω) 3

4 Step 2. Refining the estimate obtained in Step 1 using (2) with R W (h) ande 2 W estimated from the observations on {W t }. ˆf X (ω) = 1 Ê 2 W ˆf Y (ω) 1 Ê 2 W ˆf Y (ω) ˆl W (ω), where ˆl W (ω) the Fourier transformation of ˆR W (h) ˆR W (h)+ê 2 W Remark. (a) The spectral density estimation of {Y t } isanonmissingdataproblem, hence the classical spectral estimation methods can be used in Step 1. (b) Due to the sampling variability, refining the spectral estimate using (2) may produce negative values. If this happens, the value zero should be used. Note if we assume that {W t } is IID, then we have VW 2 f W (λ) = (VW 2 + E2 W ). Hence (2) becomes F X (λ) = 1 V EW 2 W 2 F Y (λ) V Y 2 MW 2 (V W 2 + (4) E2 W ), where VW 2 and V Y 2 are the variances of {W t}, respectively. Refining spectral estimates using (4) will be very simple. 3 Evaluation of the Approach In this section we use simulated data to compare the two-step approach described in Section 2 to the classical periodogram, the Lomb periodogram and the svd periodogram. Although the two-step approach can (should) be used with smoothed periodogram estimators, for fair comparison (since the classical periodogram, the Lomb periodogram and the svd periodogram are not involed smoothing), we use the periodogram in Step 1. In such case the approachcanbecalledtherefined periodogram. The simulated data are generated by the followin two steps: Step 1. The S-PLUS function arima.sim is used to generate data {x 1,..., x T } with T =128 from the AR(2) model X t = a 1 X t 1 + a 2 X t 2 + Z t, where a 1 =0.4, a 2 = 0.8 and{z t } are IID N(0,1). The spectral density of the model is f(λ) = 1 2π{(1 a 2 ) 2 + a a 1(1 + a 2 )cos(λ)+4a 2 cos 2 (λ)} 4

5 This density function has a peak at the medium frequency (see Figure 1). Step 2. The missing data are generated by deleting x t from the data set if u t > 0.8, for t=1,...,t. Where {u 1,..,u T } are a random sample from the uniform distribution U[0,1]. The refined periodogram is calculated using the simulated data. To use (4), E W is estimated as Ê W = the number of missings divided by T,andVW 2 is estimated as ˆV W 2 = Ê W (1 Ê W ). As comparison the classical periodogram, the SVD periodogram and the Lomb periodogram are computed using the simulated data. This process is repeated 100 times. The sample mean and variance of each estimate are calculated and recorded. Figure 1 gives the sample mean of each spectral estimate. In this figure we can see that the sample mean of the refined periodogram is much closer to the true spectrum than the ones of the other methods. Hence the refined periodogram substantially reduced the estimation bias. Figure 1 also show that the sample means of the SVD periodogram and the Lomb periodogram and the classical periodogram are very close to each other. It seems that the estimation bias cannot bt improved by the Lomb periodogram and the SVD periodogram. In fact Scargle (1982) pointed out that for the Lomb periodogram and the classical periodogram the actual values are typically not changed much, even though the form is significantly changed. Figure 2 shows the sample variance of each spectral estimate. In this figure we can see that the refined periodogram has larger variance than the SVD periodogram and the Lomb periodogram and the classical periodogram. Figure 2 also show there no big difference among the SVD periodogram and the Lomb periodogram and the classical periodogram. Comparing to the other periodograms, the refined periodogram has the smaller bias but larger variance. Since the standard deviations of the spectral estimate are (approximately) proportional to the power spectrum, Parzen (1957) suggested to use mean squared percentage error to measure the estimation accuracy, which is defined by K(ω) ={v(ω) 2 + b(ω) 2 }/f(ω) 2 In this simulation study, b(w) iscomputedasthesamplemean-f(ω), and v(w) 2 is taken as the sample variance. Figure 3 shows the mean squared percentage error of each spectral estimate. From Figure 3, we can see that the refined periodogram has smaller MSPE than the other periodograms 5

6 Power spectrum Classical Modified Lomb SVD Frequency Figure 1: The sample mean of of the spectral estimates. The results based on 100 simulations with T=128 and 20% missing Modified Classical Lomb SVD Frequency Figure 2: The variance of the spectral estimates. The results based on 100 simulation with T=128 and 20% missing. 6

7 percentage error Modified Classical Lomb SVD Frequency Figure 3: The mean squared percentage error of the spectral estimates. The results based on 100 simulations with T=128 and 20% missing. at most frequencies. The average value of MSPE over the frequencies is reported in Table 1 Theaveragevaluesofthepercentagebiassquired(PBS)andthepercentagevariance(PV) are also reported in Table 1. Table 1. The average values of mean squared percentage error and percentage bias squared and the percentage variance over the frequencies. The results are based on 100 simulations with T=128 and 20% missing. Periodograms PBS PV MSPE(SE) Classical (0.11) Lomb (0.11) SVD (0.11) Refined (0.09) 4 Application to a real data set The rainfall data in Ireland consist of monthly average daily rainfall records at 1294 stations over the period of 37 years. Data missing patterns are quite different among the stations. As an example to illustrate the application of our technique we select rainfall data in 7

8 Figure 4: The Monthly average daily rainfall anomaly at DURROW G.S. DURROW G.S. (7 o , 52 o ). Annual rainfall pattern are defind as the monthly (daily) rainfall amount averaged over 37 years. The anomaly are defined by subtracting annual rainfall pattern from the data and shown in Figure 4. Figure 4 seems show that the data have randomly (26Formula (4) is used to compute the refined periodogram with Ê W =0.74 and ˆV W 2 = the refined periodogram of the anormaly is shown in Figure 5. Figure 5 seems indicate that the anormaly are white noise. 5 Conclusion and discussion We have investigated a new approach for estimating the power spectrum of stationary time series with missing data. The new approach is the refined spectral estimate of non missing data, its asymptotical unbiasness is easy to be established. This property does not hold for the classical periodogram and the Lomb periodogram and the SVD periodogram for missing data. The simulation results show that the new approach substantially reduced the bias and slightly increased the variance (relative to the other periodogram methods). Overall the new approach significantly reduced the mean squared percentage error. The simulation results also show that the actual values of the classical periodogram and the Lomb periodogram and the SVD periodogram are quite similar. 8

9 Frequency Figure 5: The refined periodogram of the anomaly in Figure 4 In climatological studies, data missing is usually time dependent. For example we have more missing in winter than in summer in rainfall data in Ireland. Hence further work can be focused on smoothed periodogram estimators and non stationary weights {W t }. References Brockwell PJ and Davis RA (1993). Time series: theory and methods. Springer-Verlag, New York. Fortin P and Mackey MC (1999). Periodic chronic myelogenous leukaemia: spectral analysis of blood cell counts and aetiological implications. BritishJ.ofHaematology, 102(2), Green D, Maclaren DA, Allison W, and Dastoor PC. (2002). Developments in the analysis of helium atom scattering data: the sequential filtering Lomb periodogram technique. J. of Physics D-Applied Physics 24, Kaneoke Y and Vitek JL (1996) Burst and oscillations as disparate neuronal properties. J. of Neuroscience Methods, 68(2),

10 Laguna P, Moody GB, Mark GR (1989) Power spectral density of unevenly sampled data by least-square analysis: Performance and application to heart rate signals. IEEE TransactiononiBiomedicalEngineering, 45, Lomb NR (1996) Least squares frequency analysis of unevenly spaced data. Astrophysical and space science, 39, Parzen E (1957) On choosing an estimate of the spectral density function of a stationary time series. Ann. Math. Statist Priestley MB (1996) Spectral analysis and time series. Academic Press, New York. Scargle, JD (1989) Studies in astronomical time series analysis II. Statistical aspects of spectral analysis of unevenly sampled data. Astrophysical Journal, 263,

Time Series. Anthony Davison. c

Time Series. Anthony Davison. c Series Anthony Davison c 2008 http://stat.epfl.ch Periodogram 76 Motivation............................................................ 77 Lutenizing hormone data..................................................

More information

Time Series Analysis -- An Introduction -- AMS 586

Time Series Analysis -- An Introduction -- AMS 586 Time Series Analysis -- An Introduction -- AMS 586 1 Objectives of time series analysis Data description Data interpretation Modeling Control Prediction & Forecasting 2 Time-Series Data Numerical data

More information

Fourier Analysis of Stationary and Non-Stationary Time Series

Fourier Analysis of Stationary and Non-Stationary Time Series Fourier Analysis of Stationary and Non-Stationary Time Series September 6, 2012 A time series is a stochastic process indexed at discrete points in time i.e X t for t = 0, 1, 2, 3,... The mean is defined

More information

8.2 Harmonic Regression and the Periodogram

8.2 Harmonic Regression and the Periodogram Chapter 8 Spectral Methods 8.1 Introduction Spectral methods are based on thining of a time series as a superposition of sinusoidal fluctuations of various frequencies the analogue for a random process

More information

Estimating Periodic Signals

Estimating Periodic Signals Department of Mathematics & Statistics Indian Institute of Technology Kanpur Most of this talk has been taken from the book Statistical Signal Processing, by D. Kundu and S. Nandi. August 26, 2012 Outline

More information

Frequency estimation by DFT interpolation: A comparison of methods

Frequency estimation by DFT interpolation: A comparison of methods Frequency estimation by DFT interpolation: A comparison of methods Bernd Bischl, Uwe Ligges, Claus Weihs March 5, 009 Abstract This article comments on a frequency estimator which was proposed by [6] and

More information

SF2943: TIME SERIES ANALYSIS COMMENTS ON SPECTRAL DENSITIES

SF2943: TIME SERIES ANALYSIS COMMENTS ON SPECTRAL DENSITIES SF2943: TIME SERIES ANALYSIS COMMENTS ON SPECTRAL DENSITIES This document is meant as a complement to Chapter 4 in the textbook, the aim being to get a basic understanding of spectral densities through

More information

6.435, System Identification

6.435, System Identification System Identification 6.435 SET 3 Nonparametric Identification Munther A. Dahleh 1 Nonparametric Methods for System ID Time domain methods Impulse response Step response Correlation analysis / time Frequency

More information

STAT 520: Forecasting and Time Series. David B. Hitchcock University of South Carolina Department of Statistics

STAT 520: Forecasting and Time Series. David B. Hitchcock University of South Carolina Department of Statistics David B. University of South Carolina Department of Statistics What are Time Series Data? Time series data are collected sequentially over time. Some common examples include: 1. Meteorological data (temperatures,

More information

Time Series Analysis. Solutions to problems in Chapter 7 IMM

Time Series Analysis. Solutions to problems in Chapter 7 IMM Time Series Analysis Solutions to problems in Chapter 7 I Solution 7.1 Question 1. As we get by subtraction kx t = 1 + B + B +...B k 1 )ǫ t θb) = 1 + B +... + B k 1 ), and BθB) = B + B +... + B k ), 1

More information

Unstable Oscillations!

Unstable Oscillations! Unstable Oscillations X( t ) = [ A 0 + A( t ) ] sin( ω t + Φ 0 + Φ( t ) ) Amplitude modulation: A( t ) Phase modulation: Φ( t ) S(ω) S(ω) Special case: C(ω) Unstable oscillation has a broader periodogram

More information

Non-parametric identification

Non-parametric identification Non-parametric Non-parametric Transient Step-response using Spectral Transient Correlation Frequency function estimate Spectral System Identification, SSY230 Non-parametric 1 Non-parametric Transient Step-response

More information

Statistical Applications in the Astronomy Literature II Jogesh Babu. Center for Astrostatistics PennState University, USA

Statistical Applications in the Astronomy Literature II Jogesh Babu. Center for Astrostatistics PennState University, USA Statistical Applications in the Astronomy Literature II Jogesh Babu Center for Astrostatistics PennState University, USA 1 The likelihood ratio test (LRT) and the related F-test Protassov et al. (2002,

More information

Introduction to Spectral and Time-Spectral Analysis with some Applications

Introduction to Spectral and Time-Spectral Analysis with some Applications Introduction to Spectral and Time-Spectral Analysis with some Applications A.INTRODUCTION Consider the following process X t = U cos[(=3)t] + V sin[(=3)t] Where *E[U] = E[V ] = 0 *E[UV ] = 0 *V ar(u) =

More information

Time Series: Theory and Methods

Time Series: Theory and Methods Peter J. Brockwell Richard A. Davis Time Series: Theory and Methods Second Edition With 124 Illustrations Springer Contents Preface to the Second Edition Preface to the First Edition vn ix CHAPTER 1 Stationary

More information

arxiv:cond-mat/ v1 [cond-mat.stat-mech] 25 Feb 2003

arxiv:cond-mat/ v1 [cond-mat.stat-mech] 25 Feb 2003 arxiv:cond-mat/0302507v1 [cond-mat.stat-mech] 25 Feb 2003 SIGNIFICANCE OF LOG-PERIODIC SIGNATURES IN CUMULATIVE NOISE HANS-CHRISTIAN GRAF V. BOTHMER Abstract. Using methods introduced by Scargle we derive

More information

Applied Time Series Topics

Applied Time Series Topics Applied Time Series Topics Ivan Medovikov Brock University April 16, 2013 Ivan Medovikov, Brock University Applied Time Series Topics 1/34 Overview 1. Non-stationary data and consequences 2. Trends and

More information

Periodogram of a sinusoid + spike Single high value is sum of cosine curves all in phase at time t 0 :

Periodogram of a sinusoid + spike Single high value is sum of cosine curves all in phase at time t 0 : Periodogram of a sinusoid + spike Single high value is sum of cosine curves all in phase at time t 0 : X(t) = µ + Asin(ω 0 t)+ Δ δ ( t t 0 ) ±σ N =100 Δ =100 χ ( ω ) Raises the amplitude uniformly at all

More information

Lecture 11: Spectral Analysis

Lecture 11: Spectral Analysis Lecture 11: Spectral Analysis Methods For Estimating The Spectrum Walid Sharabati Purdue University Latest Update October 27, 2016 Professor Sharabati (Purdue University) Time Series Analysis October 27,

More information

Statistics of stochastic processes

Statistics of stochastic processes Introduction Statistics of stochastic processes Generally statistics is performed on observations y 1,..., y n assumed to be realizations of independent random variables Y 1,..., Y n. 14 settembre 2014

More information

Parameter estimation in epoch folding analysis

Parameter estimation in epoch folding analysis ASTRONOMY & ASTROPHYSICS MAY II 1996, PAGE 197 SUPPLEMENT SERIES Astron. Astrophys. Suppl. Ser. 117, 197-21 (1996) Parameter estimation in epoch folding analysis S. Larsson Stockholm Observatory, S-13336

More information

WAVEPAL. A Python software for the frequency and wavelet analyses of irregularly sampled time series. Guillaume Lenoir

WAVEPAL. A Python software for the frequency and wavelet analyses of irregularly sampled time series. Guillaume Lenoir WAVEPAL A Python software for the frequency and wavelet analyses of irregularly sampled time series Guillaume Lenoir Talk at the VUB 24/05/2018 MY RESEARCH WORLD Statistics & signal processing { Paleoclimate

More information

Autoregressive Models Fourier Analysis Wavelets

Autoregressive Models Fourier Analysis Wavelets Autoregressive Models Fourier Analysis Wavelets BFR Flood w/10yr smooth Spectrum Annual Max: Day of Water year Flood magnitude vs. timing Jain & Lall, 2000 Blacksmith Fork, Hyrum, UT Analyses of Flood

More information

Period. End of Story?

Period. End of Story? Period. End of Story? Twice Around With Period Detection: Updated Algorithms Eric L. Michelsen UCSD CASS Journal Club, January 9, 2015 Can you read this? 36 pt This is big stuff 24 pt This makes a point

More information

X random; interested in impact of X on Y. Time series analogue of regression.

X random; interested in impact of X on Y. Time series analogue of regression. Multiple time series Given: two series Y and X. Relationship between series? Possible approaches: X deterministic: regress Y on X via generalized least squares: arima.mle in SPlus or arima in R. We have

More information

Problem Set 2 Solution Sketches Time Series Analysis Spring 2010

Problem Set 2 Solution Sketches Time Series Analysis Spring 2010 Problem Set 2 Solution Sketches Time Series Analysis Spring 2010 Forecasting 1. Let X and Y be two random variables such that E(X 2 ) < and E(Y 2 )

More information

Comments on New Approaches in Period Analysis of Astronomical Time Series by Pavlos Protopapas (Or: A Pavlosian Response ) Don Percival

Comments on New Approaches in Period Analysis of Astronomical Time Series by Pavlos Protopapas (Or: A Pavlosian Response ) Don Percival Comments on New Approaches in Period Analysis of Astronomical Time Series by Pavlos Protopapas (Or: A Pavlosian Response ) Don Percival Applied Physics Laboratory Department of Statistics University of

More information

Time series models in the Frequency domain. The power spectrum, Spectral analysis

Time series models in the Frequency domain. The power spectrum, Spectral analysis ime series models in the Frequency domain he power spectrum, Spectral analysis Relationship between the periodogram and the autocorrelations = + = ( ) ( ˆ α ˆ ) β I Yt cos t + Yt sin t t= t= ( ( ) ) cosλ

More information

COMPUTATION OF FOURIER TRANSFORMS FOR NOISY B FOR NOISY BANDLIMITED SIGNALS

COMPUTATION OF FOURIER TRANSFORMS FOR NOISY B FOR NOISY BANDLIMITED SIGNALS COMPUTATION OF FOURIER TRANSFORMS FOR NOISY BANDLIMITED SIGNALS October 22, 2011 I. Introduction Definition of Fourier transform: F [f ](ω) := ˆf (ω) := + f (t)e iωt dt, ω R (1) I. Introduction Definition

More information

A COMPLETE ASYMPTOTIC SERIES FOR THE AUTOCOVARIANCE FUNCTION OF A LONG MEMORY PROCESS. OFFER LIEBERMAN and PETER C. B. PHILLIPS

A COMPLETE ASYMPTOTIC SERIES FOR THE AUTOCOVARIANCE FUNCTION OF A LONG MEMORY PROCESS. OFFER LIEBERMAN and PETER C. B. PHILLIPS A COMPLETE ASYMPTOTIC SERIES FOR THE AUTOCOVARIANCE FUNCTION OF A LONG MEMORY PROCESS BY OFFER LIEBERMAN and PETER C. B. PHILLIPS COWLES FOUNDATION PAPER NO. 1247 COWLES FOUNDATION FOR RESEARCH IN ECONOMICS

More information

False Alarm Probability based on bootstrap and extreme-value methods for periodogram peaks

False Alarm Probability based on bootstrap and extreme-value methods for periodogram peaks False Alarm Probability based on bootstrap and extreme-value methods for periodogram peaks M. Süveges ISDC Data Centre for Astrophysics, University of Geneva, Switzerland Abstract One of the most pertinent

More information

On Moving Average Parameter Estimation

On Moving Average Parameter Estimation On Moving Average Parameter Estimation Niclas Sandgren and Petre Stoica Contact information: niclas.sandgren@it.uu.se, tel: +46 8 473392 Abstract Estimation of the autoregressive moving average (ARMA)

More information

Statistics of Stochastic Processes

Statistics of Stochastic Processes Prof. Dr. J. Franke All of Statistics 4.1 Statistics of Stochastic Processes discrete time: sequence of r.v...., X 1, X 0, X 1, X 2,... X t R d in general. Here: d = 1. continuous time: random function

More information

Local Whittle Likelihood Estimators and Tests for non-gaussian Linear Processes

Local Whittle Likelihood Estimators and Tests for non-gaussian Linear Processes Local Whittle Likelihood Estimators and Tests for non-gaussian Linear Processes By Tomohito NAITO, Kohei ASAI and Masanobu TANIGUCHI Department of Mathematical Sciences, School of Science and Engineering,

More information

Step Response Analysis. Frequency Response, Relation Between Model Descriptions

Step Response Analysis. Frequency Response, Relation Between Model Descriptions Step Response Analysis. Frequency Response, Relation Between Model Descriptions Automatic Control, Basic Course, Lecture 3 November 9, 27 Lund University, Department of Automatic Control Content. Step

More information

ECO 513 Fall 2009 C. Sims CONDITIONAL EXPECTATION; STOCHASTIC PROCESSES

ECO 513 Fall 2009 C. Sims CONDITIONAL EXPECTATION; STOCHASTIC PROCESSES ECO 513 Fall 2009 C. Sims CONDIIONAL EXPECAION; SOCHASIC PROCESSES 1. HREE EXAMPLES OF SOCHASIC PROCESSES (I) X t has three possible time paths. With probability.5 X t t, with probability.25 X t t, and

More information

Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification

Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification Page 1 of 13 Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification The use of the Lomb-Scargle Periodogram (LSP) for the analysis of biological signal rhythms

More information

Periodogram of a sinusoid + spike Single high value is sum of cosine curves all in phase at time t 0 :

Periodogram of a sinusoid + spike Single high value is sum of cosine curves all in phase at time t 0 : Periodogram of a sinusoid + spike Single high value is sum of cosine curves all in phase at time t 0 : ( ) ±σ X(t) = µ + Asin(ω 0 t)+ Δ δ t t 0 N =100 Δ =100 χ ( ω ) Raises the amplitude uniformly at all

More information

Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science : Discrete-Time Signal Processing

Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science : Discrete-Time Signal Processing Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science 6.34: Discrete-Time Signal Processing OpenCourseWare 006 ecture 8 Periodogram Reading: Sections 0.6 and 0.7

More information

Gaussian processes. Basic Properties VAG002-

Gaussian processes. Basic Properties VAG002- Gaussian processes The class of Gaussian processes is one of the most widely used families of stochastic processes for modeling dependent data observed over time, or space, or time and space. The popularity

More information

LIGHT CURVE PROPERTIES OF HIGH EARTH OBITAL ROCKET BODIES OBSERVED AT WEIHAI OBSERVATORY OF SHANDONG UNIVERSITY

LIGHT CURVE PROPERTIES OF HIGH EARTH OBITAL ROCKET BODIES OBSERVED AT WEIHAI OBSERVATORY OF SHANDONG UNIVERSITY LIGHT CURVE PROPERTIES OF HIGH EARTH OBITAL ROCKET BODIES OBSERVED AT WEIHAI OBSERVATORY OF SHANDONG UNIVERSITY Xu Chen, Shao-Ming Hu, Jun-Ju Du, and Di-Fu Guo Shandong University, Weihai, 264209 Weihai,

More information

Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification

Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification Page 1 of 16 Investigating the use of the Lomb-Scargle Periodogram for Heart Rate Variability Quantification The use of the Lomb-Scargle Periodogram (LSP) for the analysis of biological signal rhythms

More information

Spectral Analysis of Irregularly Sampled Data with Time Series Models

Spectral Analysis of Irregularly Sampled Data with Time Series Models The Open Signal Processing Journal, 2008, 1, 7-14 7 Open Access Spectral Analysis of Irregularly Sampled Data with Time Series Models Piet M.T. Broersen* Department of Multi Scale Physics, Delft University

More information

sine wave fit algorithm

sine wave fit algorithm TECHNICAL REPORT IR-S3-SB-9 1 Properties of the IEEE-STD-57 four parameter sine wave fit algorithm Peter Händel, Senior Member, IEEE Abstract The IEEE Standard 57 (IEEE-STD-57) provides algorithms for

More information

EEM 409. Random Signals. Problem Set-2: (Power Spectral Density, LTI Systems with Random Inputs) Problem 1: Problem 2:

EEM 409. Random Signals. Problem Set-2: (Power Spectral Density, LTI Systems with Random Inputs) Problem 1: Problem 2: EEM 409 Random Signals Problem Set-2: (Power Spectral Density, LTI Systems with Random Inputs) Problem 1: Consider a random process of the form = + Problem 2: X(t) = b cos(2π t + ), where b is a constant,

More information

Principal Dynamical Components

Principal Dynamical Components Principal Dynamical Components Manuel D. de la Iglesia* Departamento de Análisis Matemático, Universidad de Sevilla Instituto Nacional de Matemática Pura e Aplicada (IMPA) Rio de Janeiro, May 14, 2013

More information

Spectral Analysis of Random Processes

Spectral Analysis of Random Processes Spectral Analysis of Random Processes Spectral Analysis of Random Processes Generally, all properties of a random process should be defined by averaging over the ensemble of realizations. Generally, all

More information

Regularization methods for large-scale, ill-posed, linear, discrete, inverse problems

Regularization methods for large-scale, ill-posed, linear, discrete, inverse problems Regularization methods for large-scale, ill-posed, linear, discrete, inverse problems Silvia Gazzola Dipartimento di Matematica - Università di Padova January 10, 2012 Seminario ex-studenti 2 Silvia Gazzola

More information

WAVELET TRANSFORMS IN TIME SERIES ANALYSIS

WAVELET TRANSFORMS IN TIME SERIES ANALYSIS WAVELET TRANSFORMS IN TIME SERIES ANALYSIS R.C. SINGH 1 Abstract The existing methods based on statistical techniques for long range forecasts of Indian summer monsoon rainfall have shown reasonably accurate

More information

A time series is called strictly stationary if the joint distribution of every collection (Y t

A time series is called strictly stationary if the joint distribution of every collection (Y t 5 Time series A time series is a set of observations recorded over time. You can think for example at the GDP of a country over the years (or quarters) or the hourly measurements of temperature over a

More information

ESTIMATION OF PARAMETERS OF PARTIALLY SINUSOIDAL FREQUENCY MODEL

ESTIMATION OF PARAMETERS OF PARTIALLY SINUSOIDAL FREQUENCY MODEL 1 ESTIMATION OF PARAMETERS OF PARTIALLY SINUSOIDAL FREQUENCY MODEL SWAGATA NANDI 1 AND DEBASIS KUNDU Abstract. In this paper, we propose a modification of the multiple sinusoidal model such that periodic

More information

covariance function, 174 probability structure of; Yule-Walker equations, 174 Moving average process, fluctuations, 5-6, 175 probability structure of

covariance function, 174 probability structure of; Yule-Walker equations, 174 Moving average process, fluctuations, 5-6, 175 probability structure of Index* The Statistical Analysis of Time Series by T. W. Anderson Copyright 1971 John Wiley & Sons, Inc. Aliasing, 387-388 Autoregressive {continued) Amplitude, 4, 94 case of first-order, 174 Associated

More information

Problem Sheet 1 Examples of Random Processes

Problem Sheet 1 Examples of Random Processes RANDOM'PROCESSES'AND'TIME'SERIES'ANALYSIS.'PART'II:'RANDOM'PROCESSES' '''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''Problem'Sheets' Problem Sheet 1 Examples of Random Processes 1. Give

More information

Dr. David A. Clifton Group Leader Computational Health Informatics (CHI) Lab Lecturer in Engineering Science, Balliol College

Dr. David A. Clifton Group Leader Computational Health Informatics (CHI) Lab Lecturer in Engineering Science, Balliol College Dr. David A. Clifton Group Leader Computational Health Informatics (CHI) Lab Lecturer in Engineering Science, Balliol College 1. Introduction to Fourier analysis, the Fourier series 2. Sampling and Aliasing

More information

13.7 Power Spectrum Estimation by the Maximum Entropy (All Poles) Method

13.7 Power Spectrum Estimation by the Maximum Entropy (All Poles) Method 3.7 Maximum Entropy (All Poles) Method 565 LP coefficients for each segment of the data. The output is reconstructed by driving these coefficients with initial conditions consisting of all zeros except

More information

Introduction to Biomedical Engineering

Introduction to Biomedical Engineering Introduction to Biomedical Engineering Biosignal processing Kung-Bin Sung 6/11/2007 1 Outline Chapter 10: Biosignal processing Characteristics of biosignals Frequency domain representation and analysis

More information

Modeling and testing long memory in random fields

Modeling and testing long memory in random fields Modeling and testing long memory in random fields Frédéric Lavancier lavancier@math.univ-lille1.fr Université Lille 1 LS-CREST Paris 24 janvier 6 1 Introduction Long memory random fields Motivations Previous

More information

SPECTRAL ANALYSIS OF NON-UNIFORMLY SAMPLED DATA: A NEW APPROACH VERSUS THE PERIODOGRAM

SPECTRAL ANALYSIS OF NON-UNIFORMLY SAMPLED DATA: A NEW APPROACH VERSUS THE PERIODOGRAM SPECTRAL ANALYSIS OF NON-UNIFORMLY SAMPLED DATA: A NEW APPROACH VERSUS THE PERIODOGRAM Hao He, Jian Li and Petre Stoica Dept. of Electrical and Computer Engineering, University of Florida, Gainesville,

More information

Time Series Examples Sheet

Time Series Examples Sheet Lent Term 2001 Richard Weber Time Series Examples Sheet This is the examples sheet for the M. Phil. course in Time Series. A copy can be found at: http://www.statslab.cam.ac.uk/~rrw1/timeseries/ Throughout,

More information

Computational Data Analysis!

Computational Data Analysis! 12.714 Computational Data Analysis! Alan Chave (alan@whoi.edu)! Thomas Herring (tah@mit.edu),! http://geoweb.mit.edu/~tah/12.714! Introduction to Spectral Analysis! Topics Today! Aspects of Time series

More information

The Caterpillar -SSA approach: automatic trend extraction and other applications. Theodore Alexandrov

The Caterpillar -SSA approach: automatic trend extraction and other applications. Theodore Alexandrov The Caterpillar -SSA approach: automatic trend extraction and other applications Theodore Alexandrov theo@pdmi.ras.ru St.Petersburg State University, Russia Bremen University, 28 Feb 2006 AutoSSA: http://www.pdmi.ras.ru/

More information

Chapter 6: Nonparametric Time- and Frequency-Domain Methods. Problems presented by Uwe

Chapter 6: Nonparametric Time- and Frequency-Domain Methods. Problems presented by Uwe System Identification written by L. Ljung, Prentice Hall PTR, 1999 Chapter 6: Nonparametric Time- and Frequency-Domain Methods Problems presented by Uwe System Identification Problems Chapter 6 p. 1/33

More information

ECON 616: Lecture 1: Time Series Basics

ECON 616: Lecture 1: Time Series Basics ECON 616: Lecture 1: Time Series Basics ED HERBST August 30, 2017 References Overview: Chapters 1-3 from Hamilton (1994). Technical Details: Chapters 2-3 from Brockwell and Davis (1987). Intuition: Chapters

More information

Stochastic Processes. M. Sami Fadali Professor of Electrical Engineering University of Nevada, Reno

Stochastic Processes. M. Sami Fadali Professor of Electrical Engineering University of Nevada, Reno Stochastic Processes M. Sami Fadali Professor of Electrical Engineering University of Nevada, Reno 1 Outline Stochastic (random) processes. Autocorrelation. Crosscorrelation. Spectral density function.

More information

observational data analysis (Supplementary Material)

observational data analysis (Supplementary Material) 1 Assessing the equatorial long-wave approximation: asymptotics and observational data analysis (Supplementary Material) 3 by H. Reed Ogrosky and Samuel N. Stechmann 4 1. Kelvin and MRG waves 5 6 The analysis

More information

Frequencies in Ultra-high-frequent Trading

Frequencies in Ultra-high-frequent Trading Frequencies in Ultra-high-frequent Trading Wing Lon Ng University of Muenster Institute of Econometrics and Economic Statistics E-mail: 05wing@wiwi.uni muenster.de January 8, 2007 Abstract This paper suggests

More information

Spectral Analysis - Introduction

Spectral Analysis - Introduction Spectral Analysis - Introduction Initial Code Introduction Summer Semester 011/01 Lukas Vacha In previous lectures we studied stationary time series in terms of quantities that are functions of time. The

More information

Notes on the Periodically Forced Harmonic Oscillator

Notes on the Periodically Forced Harmonic Oscillator Notes on the Periodically orced Harmonic Oscillator Warren Weckesser Math 38 - Differential Equations 1 The Periodically orced Harmonic Oscillator. By periodically forced harmonic oscillator, we mean the

More information

Made available courtesy of Elsevier:

Made available courtesy of Elsevier: A note on processes with random stationary increments By: Haimeng Zhang, Chunfeng Huang Zhang, H. and Huang, C. (2014). A Note on Processes with Random Stationary Increments. Statistics and Probability

More information

Notes on Fourier Series and Integrals Fourier Series

Notes on Fourier Series and Integrals Fourier Series Notes on Fourier Series and Integrals Fourier Series et f(x) be a piecewise linear function on [, ] (This means that f(x) may possess a finite number of finite discontinuities on the interval). Then f(x)

More information

Class: Trend-Cycle Decomposition

Class: Trend-Cycle Decomposition Class: Trend-Cycle Decomposition Macroeconometrics - Spring 2011 Jacek Suda, BdF and PSE June 1, 2011 Outline Outline: 1 Unobserved Component Approach 2 Beveridge-Nelson Decomposition 3 Spectral Analysis

More information

Using wavelet tools to estimate and assess trends in atmospheric data

Using wavelet tools to estimate and assess trends in atmospheric data NRCSE Using wavelet tools to estimate and assess trends in atmospheric data Wavelets Fourier analysis uses big waves Wavelets are small waves Requirements for wavelets Integrate to zero Square integrate

More information

data lam=36.9 lam=6.69 lam=4.18 lam=2.92 lam=2.21 time max wavelength modulus of max wavelength cycle

data lam=36.9 lam=6.69 lam=4.18 lam=2.92 lam=2.21 time max wavelength modulus of max wavelength cycle AUTOREGRESSIVE LINEAR MODELS AR(1) MODELS The zero-mean AR(1) model x t = x t,1 + t is a linear regression of the current value of the time series on the previous value. For > 0 it generates positively

More information

4. Complex Oscillations

4. Complex Oscillations 4. Complex Oscillations The most common use of complex numbers in physics is for analyzing oscillations and waves. We will illustrate this with a simple but crucially important model, the damped harmonic

More information

ARIMA modeling to forecast area and production of rice in West Bengal

ARIMA modeling to forecast area and production of rice in West Bengal Journal of Crop and Weed, 9(2):26-31(2013) ARIMA modeling to forecast area and production of rice in West Bengal R. BISWAS AND B. BHATTACHARYYA Department of Agricultural Statistics Bidhan Chandra Krishi

More information

Part III Example Sheet 1 - Solutions YC/Lent 2015 Comments and corrections should be ed to

Part III Example Sheet 1 - Solutions YC/Lent 2015 Comments and corrections should be  ed to TIME SERIES Part III Example Sheet 1 - Solutions YC/Lent 2015 Comments and corrections should be emailed to Y.Chen@statslab.cam.ac.uk. 1. Let {X t } be a weakly stationary process with mean zero and let

More information

Energy during a burst of deceleration

Energy during a burst of deceleration Problem 1. Energy during a burst of deceleration A particle of charge e moves at constant velocity, βc, for t < 0. During the short time interval, 0 < t < t its velocity remains in the same direction but

More information

Prof. Dr. Roland Füss Lecture Series in Applied Econometrics Summer Term Introduction to Time Series Analysis

Prof. Dr. Roland Füss Lecture Series in Applied Econometrics Summer Term Introduction to Time Series Analysis Introduction to Time Series Analysis 1 Contents: I. Basics of Time Series Analysis... 4 I.1 Stationarity... 5 I.2 Autocorrelation Function... 9 I.3 Partial Autocorrelation Function (PACF)... 14 I.4 Transformation

More information

Chapter 2 Random Processes

Chapter 2 Random Processes Chapter 2 Random Processes 21 Introduction We saw in Section 111 on page 10 that many systems are best studied using the concept of random variables where the outcome of a random experiment was associated

More information

A Comparison of HRV Techniques: The Lomb Periodogram versus The Smoothed Pseudo Wigner-Ville Distribution

A Comparison of HRV Techniques: The Lomb Periodogram versus The Smoothed Pseudo Wigner-Ville Distribution A Comparison of HRV Techniques: The Lomb Periodogram versus The Smoothed Pseudo Wigner-Ville Distribution By: Mark Ebden Submitted to: Prof. Lionel Tarassenko Date: 19 November, 2002 (Revised 20 November)

More information

Introduction to Signal Processing

Introduction to Signal Processing to Signal Processing Davide Bacciu Dipartimento di Informatica Università di Pisa bacciu@di.unipi.it Intelligent Systems for Pattern Recognition Signals = Time series Definitions Motivations A sequence

More information

Time Series Prediction

Time Series Prediction Sternberg Astronomical Institute Lomonosov Moscow State University Time Series Prediction Leonid Zotov Fulbright Scholar 28-29 San Juan, Puerto Rico, 8 May 29 FROM THE BEGINNING OF THE HISTORY PEOPLE TRIED

More information

The distribution inherited by Y is called the Cauchy distribution. Using that. d dy ln(1 + y2 ) = 1 arctan(y)

The distribution inherited by Y is called the Cauchy distribution. Using that. d dy ln(1 + y2 ) = 1 arctan(y) Stochastic Processes - MM3 - Solutions MM3 - Review Exercise Let X N (0, ), i.e. X is a standard Gaussian/normal random variable, and denote by f X the pdf of X. Consider also a continuous random variable

More information

Statistics 910, #5 1. Regression Methods

Statistics 910, #5 1. Regression Methods Statistics 910, #5 1 Overview Regression Methods 1. Idea: effects of dependence 2. Examples of estimation (in R) 3. Review of regression 4. Comparisons and relative efficiencies Idea Decomposition Well-known

More information

II. Nonparametric Spectrum Estimation for Stationary Random Signals - Non-parametric Methods -

II. Nonparametric Spectrum Estimation for Stationary Random Signals - Non-parametric Methods - II. onparametric Spectrum Estimation for Stationary Random Signals - on-parametric Methods - - [p. 3] Periodogram - [p. 12] Periodogram properties - [p. 23] Modified periodogram - [p. 25] Bartlett s method

More information

REGULARIZED ESTIMATION OF LINE SPECTRA FROM IRREGULARLY SAMPLED ASTROPHYSICAL DATA. Sébastien Bourguignon, Hervé Carfantan, Loïc Jahan

REGULARIZED ESTIMATION OF LINE SPECTRA FROM IRREGULARLY SAMPLED ASTROPHYSICAL DATA. Sébastien Bourguignon, Hervé Carfantan, Loïc Jahan REGULARIZED ESTIMATION OF LINE SPECTRA FROM IRREGULARLY SAMPLED ASTROPHYSICAL DATA Sébastien Bourguignon, Hervé Carfantan, Loïc Jahan Laboratoire d Astrophysique de Toulouse - Tarbes, UMR 557 CNRS/UPS

More information

A NEW APPROACH TO SPECTRAL ESTIMATION FROM IRREGULAR SAMPLING. David BONACCI and Bernard LACAZE

A NEW APPROACH TO SPECTRAL ESTIMATION FROM IRREGULAR SAMPLING. David BONACCI and Bernard LACAZE A NEW APPROACH TO SPECTRAL ESTIMATION FROM IRREGULAR SAMPLING David BONACCI and Bernard LACAZE TESA Laboratory Signal Processing Unit 1- Port Saint-Etienne - 31 TOULOUSE - FRANCE ABSTRACT This article

More information

Scientific Computing: An Introductory Survey

Scientific Computing: An Introductory Survey Scientific Computing: An Introductory Survey Chapter 12 Prof. Michael T. Heath Department of Computer Science University of Illinois at Urbana-Champaign Copyright c 2002. Reproduction permitted for noncommercial,

More information

Constructing Approximation Kernels for Non-Harmonic Fourier Data

Constructing Approximation Kernels for Non-Harmonic Fourier Data Constructing Approximation Kernels for Non-Harmonic Fourier Data Aditya Viswanathan aditya.v@caltech.edu California Institute of Technology SIAM Annual Meeting 2013 July 10 2013 0 / 19 Joint work with

More information

The effects of model misspecification on linear regression coefficients as applicable to solar and linear terms

The effects of model misspecification on linear regression coefficients as applicable to solar and linear terms The effects of model misspecification on linear regression coefficients as applicable to solar and linear terms Troy A. Wynn and Dr. Vincent B. Wickwar Utah State University, Department of Physics Abstract

More information

Fast and Robust Phase Retrieval

Fast and Robust Phase Retrieval Fast and Robust Phase Retrieval Aditya Viswanathan aditya@math.msu.edu CCAM Lunch Seminar Purdue University April 18 2014 0 / 27 Joint work with Yang Wang Mark Iwen Research supported in part by National

More information

Correlator I. Basics. Chapter Introduction. 8.2 Digitization Sampling. D. Anish Roshi

Correlator I. Basics. Chapter Introduction. 8.2 Digitization Sampling. D. Anish Roshi Chapter 8 Correlator I. Basics D. Anish Roshi 8.1 Introduction A radio interferometer measures the mutual coherence function of the electric field due to a given source brightness distribution in the sky.

More information

SF2943: A FEW COMMENTS ON SPECTRAL DENSITIES

SF2943: A FEW COMMENTS ON SPECTRAL DENSITIES SF2943: A FEW COMMENTS ON SPECTRAL DENSITIES FILIP LINDSKOG The aim of this brief note is to get some basic understanding of spectral densities by computing, plotting and interpreting the spectral densities

More information

CCNY. BME I5100: Biomedical Signal Processing. Stochastic Processes. Lucas C. Parra Biomedical Engineering Department City College of New York

CCNY. BME I5100: Biomedical Signal Processing. Stochastic Processes. Lucas C. Parra Biomedical Engineering Department City College of New York BME I5100: Biomedical Signal Processing Stochastic Processes Lucas C. Parra Biomedical Engineering Department CCNY 1 Schedule Week 1: Introduction Linear, stationary, normal - the stuff biology is not

More information

Model Selection for Geostatistical Models

Model Selection for Geostatistical Models Model Selection for Geostatistical Models Richard A. Davis Colorado State University http://www.stat.colostate.edu/~rdavis/lectures Joint work with: Jennifer A. Hoeting, Colorado State University Andrew

More information

Generalised AR and MA Models and Applications

Generalised AR and MA Models and Applications Chapter 3 Generalised AR and MA Models and Applications 3.1 Generalised Autoregressive Processes Consider an AR1) process given by 1 αb)x t = Z t ; α < 1. In this case, the acf is, ρ k = α k for k 0 and

More information

First-order random coefficient autoregressive (RCA(1)) model: Joint Whittle estimation and information

First-order random coefficient autoregressive (RCA(1)) model: Joint Whittle estimation and information ACTA ET COMMENTATIONES UNIVERSITATIS TARTUENSIS DE MATHEMATICA Volume 9, Number, June 05 Available online at http://acutm.math.ut.ee First-order random coefficient autoregressive RCA model: Joint Whittle

More information

Damped harmonic motion

Damped harmonic motion Damped harmonic motion March 3, 016 Harmonic motion is studied in the presence of a damping force proportional to the velocity. The complex method is introduced, and the different cases of under-damping,

More information

D.S.G. POLLOCK: BRIEF NOTES

D.S.G. POLLOCK: BRIEF NOTES BIVARIATE SPECTRAL ANALYSIS Let x(t) and y(t) be two stationary stochastic processes with E{x(t)} = E{y(t)} =. These processes have the following spectral representations: (1) x(t) = y(t) = {cos(ωt)da

More information

Publications (in chronological order)

Publications (in chronological order) Publications (in chronological order) 1. A note on the investigation of the optimal weight function in estimation of the spectral density (1963), J. Univ. Gau. 14, pages 141 149. 2. On the cross periodogram

More information