Directional Statistics by K. V. Mardia & P. E. Jupp Wiley, Chichester, Errata to 1st printing

Similar documents
Notes on the Multivariate Normal and Related Topics

Hypothesis Testing. Robert L. Wolpert Department of Statistical Science Duke University, Durham, NC, USA

STAT 730 Chapter 4: Estimation

Final Examination Statistics 200C. T. Ferguson June 11, 2009

Discussion Paper No. 28

Today we will prove one result from probability that will be useful in several statistical tests. ... B1 B2 Br. Figure 23.1:

Part III. Hypothesis Testing. III.1. Log-rank Test for Right-censored Failure Time Data

Lecture 3. Inference about multivariate normal distribution

5 Applying the Fokker-Planck equation

Master s Written Examination

Central Limit Theorem ( 5.3)

SYMMETRIC UNIMODAL MODELS FOR DIRECTIONAL DATA MOTIVATED BY INVERSE STEREOGRAPHIC PROJECTION

Spherical-Homoscedastic Distributions: The Equivalency of Spherical and Normal Distributions in Classification

Cosmology & CMB. Set5: Data Analysis. Davide Maino

Dr. Maddah ENMG 617 EM Statistics 10/12/12. Nonparametric Statistics (Chapter 16, Hines)

Errata for Quantum Mechanics by Ernest Abers August 1, 2004

Math 181B Homework 1 Solution

Exam 2. Jeremy Morris. March 23, 2006

Lecture 2: Basic Concepts and Simple Comparative Experiments Montgomery: Chapter 2

Errata as of September 4, 2003 Biostatistical Methods: The Assessment of Relative Risks John M. Lachin

First Year Examination Department of Statistics, University of Florida

Quantum Mechanics: A Paradigms Approach David H. McIntyre. 27 July Corrections to the 1 st printing

Linear Models in Statistics

Statistics 203: Introduction to Regression and Analysis of Variance Course review

STT 843 Key to Homework 1 Spring 2018

A correlation coefficient for circular data

ST4241 Design and Analysis of Clinical Trials Lecture 7: N. Lecture 7: Non-parametric tests for PDG data

On corrections of classical multivariate tests for high-dimensional data. Jian-feng. Yao Université de Rennes 1, IRMAR

Predictive Distributions

APPENDICES APPENDIX A. STATISTICAL TABLES AND CHARTS 651 APPENDIX B. BIBLIOGRAPHY 677 APPENDIX C. ANSWERS TO SELECTED EXERCISES 679

Canonical Correlation Analysis of Longitudinal Data

On corrections of classical multivariate tests for high-dimensional data

Notes, March 4, 2013, R. Dudley Maximum likelihood estimation: actual or supposed

STAT 730 Chapter 5: Hypothesis Testing

Master s Written Examination

Statement: With my signature I confirm that the solutions are the product of my own work. Name: Signature:.

Optimization. The value x is called a maximizer of f and is written argmax X f. g(λx + (1 λ)y) < λg(x) + (1 λ)g(y) 0 < λ < 1; x, y X.

Errata for Instructor s Solutions Manual for Gravity, An Introduction to Einstein s General Relativity 1st printing

Quick Tour of Basic Probability Theory and Linear Algebra

Example 1 describes the results from analyzing these data for three groups and two variables contained in test file manova1.tf3.

Multivariate Statistical Analysis

Errata and Updates for ASM Exam MAS-I (First Edition) Sorted by Date

Errata and Updates for ASM Exam MAS-I (First Edition) Sorted by Page

The legacy of Sir Ronald A. Fisher. Fisher s three fundamental principles: local control, replication, and randomization.

Department of Statistics

MATH5745 Multivariate Methods Lecture 07

Session 3 The proportional odds model and the Mann-Whitney test

Spring 2012 Math 541B Exam 1

Random Vectors 1. STA442/2101 Fall See last slide for copyright information. 1 / 30

Exercises Chapter 4 Statistical Hypothesis Testing

Chapter 17: Undirected Graphical Models

Institute of Actuaries of India

Akaike criterion: Kullback-Leibler discrepancy

Spherical-Homoscedastic Distributions: The Equivalency of Spherical and Normal Distributions in Classification

Econometric Methods for Panel Data

NATIONAL BOARD FOR HIGHER MATHEMATICS. M. A. and M.Sc. Scholarship Test. September 19, Time Allowed: 150 Minutes Maximum Marks: 30

Corrections and Minor Revisions of Mathematical Methods in the Physical Sciences, third edition, by Mary L. Boas (deceased)

A Study on the Correlation of Bivariate And Trivariate Normal Models

By Bhattacharjee, Das. Published: 26 April 2018

Testing Algebraic Hypotheses

COMPARISON OF FIVE TESTS FOR THE COMMON MEAN OF SEVERAL MULTIVARIATE NORMAL POPULATIONS

Multiple Linear Regression

TAMS39 Lecture 2 Multivariate normal distribution

Gaussian Models (9/9/13)

Modeling and Inference for Measured Crystal Orientations and a Tractable Class of Symmetric Distributions for Rotations in Three Dimensions

Information geometry for bivariate distribution control

exp{ (x i) 2 i=1 n i=1 (x i a) 2 (x i ) 2 = exp{ i=1 n i=1 n 2ax i a 2 i=1

INTRINSIC MEAN ON MANIFOLDS. Abhishek Bhattacharya Project Advisor: Dr.Rabi Bhattacharya

Problem Set 4. f(a + h) = P k (h) + o( h k ). (3)

Theory of Statistics.

Vectors and Matrices Statistics with Vectors and Matrices

Generalised AR and MA Models and Applications

Multivariate and Time Series Models for Circular Data with Applications to Protein Conformational Angles

Chapter 9. Hotelling s T 2 Test. 9.1 One Sample. The one sample Hotelling s T 2 test is used to test H 0 : µ = µ 0 versus

Imputation for Missing Data under PPSWR Sampling

Statistics GIDP Ph.D. Qualifying Exam Theory Jan 11, 2016, 9:00am-1:00pm

Joint work with Nottingham colleagues Simon Preston and Michail Tsagris.

Hypothesis Testing One Sample Tests

On robust and efficient estimation of the center of. Symmetry.

ACOURSEINLARGESAMPLETHEORY by Thomas S. Ferguson 1996, Chapman & Hall

ERRATA TO REAL ANALYSIS, 2nd edition (6th and later printings) G. B. Folland

Bayesian Inference. Chapter 9. Linear models and regression

Example: Four levels of herbicide strength in an experiment on dry weight of treated plants.

STAT 401A - Statistical Methods for Research Workers

Telephone-cord instabilities in thin smectic capillaries

Tests Concerning Equicorrelation Matrices with Grouped Normal Data

The Mathematics of Maps Lecture 4. Dennis The The Mathematics of Maps Lecture 4 1/29

Chapter 4 HOMEWORK ASSIGNMENTS. 4.1 Homework #1

Data Fitting - Lecture 6

ERRATA. Interest and Prices, Princeton University Press, updated February 2, 2008

I i=1 1 I(J 1) j=1 (Y ij Ȳi ) 2. j=1 (Y j Ȳ )2 ] = 2n( is the two-sample t-test statistic.

Asymptotic Variance Formulas, Gamma Functions, and Order Statistics

Wald s theorem and the Asimov data set

ABSOLUTELY CONTINUOUS SPECTRUM OF A POLYHARMONIC OPERATOR WITH A LIMIT PERIODIC POTENTIAL IN DIMENSION TWO. V (x) = V r (x); (2)

Summary of Chapter 7 (Sections ) and Chapter 8 (Section 8.1)

Katsuhiro Sugita Faculty of Law and Letters, University of the Ryukyus. Abstract

Asymptotic Statistics-III. Changliang Zou

Chapter 7. Hypothesis Testing

CS281A/Stat241A Lecture 17

STAT 512 sp 2018 Summary Sheet

Transcription:

Directional Statistics K V Mardia & P E Jupp Wiley, Chichester, 2000 Errata to 1st printing 8 4 Insert of after development 17 3 Replace θ 1 α,, θ 1 α θ 1 α,, θ n α 19 1 = (2314) Replace θ θ i 20 7 Replace 13, 13 13, 0 22 4 = (2410) The right hand side should be c 3 1 n (x i x) 3 + O(c 5 ) 22 5 = (2411) The right hand side should be 1 2c 2 1 n (x i x) 2 + 2c4 3 1 n (x i x) 4 + O(c 6 ) 22 5 Replace m 2 = 0383 + 0491i m 2 = 0383 0030i 22 3 Replace ŝ = 0322 ŝ = 0196 32 3 = (353) Replace ψ ψ θ θ T 38 4 = (3522) Replace κ 1/2 (θ µ) κ 1/2 (θ µ) 46 14 = (3548) Replace the term φ(θ; 0, Σ) in the denominator φ(µ; 0, Σ) 47 1 = (3555) Replace e 2πix e ix 51 13 Replace e a t itµ e a t +itµ 51 6 = (3570) Replace φ p = ρ p, α p = ρ p cos µ, β p = ρ p sin µ φ p = ρ p e ipµ, α p = ρ p cos pµ, β p = ρ p sin pµ 58 5 6 Replace square-summable square-integrable 58 13 Replace square-summable square-integrable 1

67 3 = (4412) Replace = 80 1 Replace nvar R nvar ( R) 89 9 Insert (up to addition of a constant) after is 89 8 = (543) The right hand side should be 90 10 Replace n 2 log ( 1 µ 2) log ( 1 µ T ) x i ˆρ 2 = 1 1 ˆµ 2 1 + ˆρ = 1 1 ˆµ 2 1 ˆµ 2 ˆµ 91 6 4 The model with density (554) is not a special case of (552) It is symmetrical about 0 and, for large enough values of µ, it has modes at ±µ 95 7 Replace O(n 1/2 ) O(n 1 ) 95 13 Replace O(n 1 ) O(n 2 ) Replace 1999 2001 95 4 = (638) This should be w = 2n {ˆκ R log I 0 (ˆκ) } = 2n {ˆκA(ˆκ) log I 0 (ˆκ)} 95 2 The right hand side should be 2n {A(ˆκ) + ˆκA (ˆκ) A(ˆκ)} 108 12 Delete the 113 5 Replace j=1 j i 117 2 = (643) The right hand side should be k p j S j j=1 2

117 4 = (644) The right hand side should be 1 n k ( Sj S ) 2 pj j=1 123 10 = (7223) The right hand side should be 2κ(n R) + 2κ(R C) 139 11 23 This subsubsection should be moved to the end of subsection 742 on page 141 (between lines 141 11 and 141 10 ), replacing ANOVA Based on a (in line 139 11 ) A, replacing asymptotic large-sample (in line 139 8 ) high-concentration asymptotic, renumbering equations as follows: (7420) (7417) (7421) (7418) (7422) (7419) (7423) (7420) (7424) (7421) (7425) (7422) (7417) (7423) (7418) (7424) (7419) (7425) and replacing equation references as follows: 139 8 (7417) (7423) 141 19 (7423) (7420) 141 17 (7423) (7420) 138 4 This should be x i x 2 = n(1 R 2 ) 139 9 = (7419) The first equation should be d i = 1 n i n i j=1 d ij 140 12 = (7424) The first term on the right hand side should be q ( ) 2 w i g 2 Ri 3

142 14 Delete is after Thus 140 1 = (753) This equation should be S = s2 c nv c (ˆκ) + s2 s nv s (ˆκ), 143 2 Replace na(ˆκ) ni 2 (ˆκ)/I 0 (ˆκ) 143 4 6 These equations should be v c (κ) = I 0(κ) 2 + I 0 (κ)i 4 (κ) 2I 2 (κ) 2 2I 0 (κ) 2 (I 0(κ)I 3 (κ) + I 0 (κ)i 1 (κ) 2I 1 (κ)i 2 (κ)) 2 2I 0 (κ) 2 (I 0 (κ) 2 + I 0 (κ)i 2 (κ) 2I 1 (κ) 2 ) v s (κ) = (I 0(κ) I 4 (κ))(i 0 (κ) I 2 (κ)) (I 1 (κ) I 3 (κ)) 2, 2I 0 (κ)(i 0 (κ) I 2 (κ)) 143 7 Replace approximations to their variances their conditional variances 151 10 Insert Assume that there are no ties before Let s i be 151 7 4 Replace these lines d 1 = 1 n 2,, d s1 = s 1 n 2, d s1+1 = s 1 + 1 n 2 1 n 1,, d s2 = s 2 n 2 1 n 1, d s2+1 = s 2 n 2 2 n 1,, d n = n 2 n 2 n 1 n 1 Hence, from (837) and some algebraic manipulation we obtain { Un 2 1,n 2 = 1 n1 ( r i n ) 2 ( i n 1 r n(n ) } 2 1 + 1) + n + n 1 (838) nn 2 n 1 2n 1 12nn 1 4

161 15 11 Replace this paragraph The projection (911) distorts the lower hemisphere more than the upper hemisphere If the data are spread over both hemispheres then it is helpful to project the two hemispheres onto separate discs in the plane, using (911) on the upper hemisphere and the variant which replaces θ π θ on the right hand side of (911) on the lower hemisphere 162 1 4 Replace tangent vectors x T µ = 0 tangent vectors z to the sphere at µ are wrapped onto the sphere z cos ( z ) µ + sin ( z ) z (913) z where z T µ = 0 164 16 14 The equation array should be S(a) = 1 n x i a 2 = 2(1 x T a) = 2(1 R x T 0 a) (923) 164 13 Replace middle expression last line 166 2,4 Replace x 0 x (3 times) 168 19 Replace κ 2 sinh κ exp { κµ T x } κ sinh κ exp { κµ T x } 169 5 Replace the integral in the numerator 1 1 teκt (1 t 2 ) (p 3)/2 dt 170 3 = (9312) Replace I (p 1)/2 (κ) I p/2 1 (κ) 172 10 Replace (1 y 2 ) 1/2 proportional to (1 y 2 ) 1/2 173 12 Replace M p (µ 2, κ) M q (µ 1, κ) 185 4 Replace x 0 R x 0 R 185 5 Replace µ 0 µ 5

185 10 Replace (958) (956) 185 11 This should be { (x1,, x n ) : x = R, x T µ = C } 185 14 15 These lines should be Integration over { (x 1,, x n ) : x T µ = C } gives the density of C as g( C; µ, κ) = c(κ) n exp{nκ C} g( C; µ, 0), and so the conditional density of R C is g ( R, C; µ, 0) g( C; µ, 0)), (9511) 185 1 = (9513) The right hand side should be c(κ) n c(nκ R) 1 f( R, R 1,, R q ; µ, 0) 186 3 = (9514) This should be 186 9 This should be f( R 1,, R q R; κ) = f( R, R 1,, R q ; µ, 0) h n (n R) ( p ) p E Xi 2 j=1 X 2 j = 1 186 7 This should be E[X 4 i ] = E [ (Xi ) ] 4 + X j 2 j i 190 17 Delete and n t 192 7 Replace tr A κ 1 193 8 Replace a b a b 198 14 15 = (1035) This should be { I ν (κ) = (2πκ) 1 2 e κ 1 4ν2 1 + (4ν2 1)(4ν 2 } 9) 8κ 2(8κ) 2 +O(κ 3 ) 6

203 6 This should be ( 12 l(a; ±x 1,, ±x n ) = n {tr(a T) log 1 F 1, p2 )}, A 207 10 Replace O(n 1/2 ) O(n 1 ) 207 7 Replace O(n 1 ) O(n 2 ) Replace 1999 2001 214 8 = (10426) The equation should be Pr(x T µ cos δ) = 1 α 215 2 = (10431) The left hand side should be sin δ 220 13,11 Replace 1 µ T 1 µ 2 1 x T 01 x 02 220 8 Replace x T i µ i x T 0i µ i 225 10 The left hand side should be x i x 2 225 1 = (10619) This should be ( q n R i i 2 n R 2 )/(q 1)(p 1) (n q n R i i 2 )/(q 1)(p 1) 226 10 = (10621) Replace ( q n ν log R ) i ν ( q n ν log R ) i ν 226 8 Replace ν i = 2(n i 1), ν = 2(n q) ν i = (p 1)(n i 1), ν = (p 1)(n q) 232 3 Replace p 2 + 6p + 20 12(p + 4) 2p 2 + 3p + 4 6(p + 4) 232 2 Replace (p 2 + 3p + 8) 3(p + 4)(p 2 + p + 2) (4p 2 + 3p 4) 3(p + 4)(p 2 + p + 2) 7

232 1 Replace p 2 4 3(p + 4)(p 2 + p + 2)(p 2 + p + 6) 4(p 2 4) 3(p + 4)(p 2 + p + 2)(p 2 + p + 6) 233 3 Replace O(n 3/2 ) O(n 2 ) Replace 1999 2001 235 11 12 (10715) and (10716) should be w b n(p2 1) { tr( T 2 ) t 2 1 (1 t 1 ) 2 /(p 1) } (10715) 2(1 2 t 1 + ĉ 11 ) w g n(p2 1) { tr( T 2 ) t 2 p (1 t p ) 2 /(p 1) }, (10716) 2(1 2 t p + ĉ pp ) 235 14 This line should be ĉ 11 = 1 n ( ) x T 4 i t 1 ĉ pp = 1 n ( ) x T 4 i t p 235 9 8 (10717) and (10718) should be n(p 2 1) { tr( T 2 ) t 2 1 (1 t 1 ) 2 /(p 1) } 2(1 2 t 1 + ĉ 11 ) n(p 2 1) { tr( T 2 ) t 2 p (1 t p ) 2 /(p 1) } 2(1 2 t p + ĉ pp ) χ 2 (p+1)(p 2)/2 (10717) χ 2 (p+1)(p 2)/2, (10718) 239 13 The right hand side should be 1 n q n i ( ) x T 4 ij t p j=1 239 6 = (10740) The right hand side should be 1 n q n i ( ) x T 4 ij t 1 j=1 8

239 4 Replace (10740) (10739) 247 1 = (1124) Replace sin 4 (πn) in the lower line sin 4 (π/n) 248 3 Replace U n = 0398 U n = 673 248 4 Replace 10% 5% Replace 459 and 46 57 and 58 248 5 Replace accepted rejected 249 7 Replace 2(r cc r ss r cs r sc )r 1 r 2 2(r cc r ss + r cs r sc )r 1 r 2 249 8 Replace r 2 cc = 0974, r 2 cs = 0213, r 2 sc = 0152, r 2 ss = 0933 r cc = 0993, r cs = 0646, r sc = 0719, r ss = 0960 249 7 Replace 0714 0713 Replace r 2 = 664 r 2 = 187 249 5 Replace nr 2 = 664 nr 2 = 187 254 10 This should be S = 1 n x ix T i y i x T i x i y T i y i y T i = S 11 S 21 S 21 S 22, 262 8 Replace A T A = I 2 and Ab = 0 A T A = ( 1 b 2) I 2, A T b = 0 and b 1 264 16 Replace is a is an 269 4 = (1233) This should be F (x) = 1 exp( ˆκx) 269 2 This should be 270 8 Replace φ i θ i ˆκ = n 1 n (1 cos θ i ) = n 1 n ( 1 R ) φ i sin θ i 271 2 Replace 0530 0389 271 6 Replace 0953 0885 271 9 Replace 0250 0490 279 5 The left hand side should be just ˆf T 9

287 8 Replace O(n 1/2 ) O(n 1 ) 287 4 Replace O(n 1 ) O(n 2 ) Replace 1999 2001 289 5 Replace Fisher matrix matrix Fisher 293 10 Replace in in in 294 11 Replace O(n 1/2 ) O(n 1 ) 294 5 Replace p 2 (p 2 + p 2) + 2r(p r)(p 2 + 4p 20) 12r(p r)(p 2)(p 1)(p + 4)(p + 2) 2p 2 (p 1)(p + 2) r(p r)(5p 2 + 2p + 8) 6r(p r)(p 2)(p + 4) 294 4 Replace [ p 2 (p 2 + p 2) r(p r)(p 2 ] 2p + 16) 3r(p r)(p 2 + p + 2)(p 2)(p + 4) [ 4p 2 (p 1)(p + 2) r(p r)(13p 2 ] + 10p 8) 3r(p r)(p 2 + p + 2)(p 2)(p + 4) 294 3 Replace (p 2r) 2 (p 1)(p + 2) 3r(p r)(p 2)(p + 4)(p 2 + p + 2)(p 2 + p + 6) 4(p 2r) 2 (p 1)(p + 2) 3r(p r)(p 2)(p + 4)(p 2 + p + 2)(p 2 + p + 6) 294 1 Replace O(n 1 ) O(n 2 ) Replace 1999 2001 300 8 Delete If H is invertible 10

306 5 Delete a before mouse 323 7 Replace O(n 1/2 ) O(n 1 ) 323 6 Replace 1999 2001 323 3 Replace O(n 1 ) O(n 2 ) 325 13 = (1452) This line should be f(x; λ, κ) = {1+κ(λ T x+1)} exp{κ(λ T x 1)}, x S 2, λ = 1, (1452) 325 12 Replace and λ is the mean shape where λ is the shape of (µ 1, µ 2, µ 3 ) and κ = 3 (µ i µ) 2 /4σ 2 349 6 5 = (A4) Replace (4p2 1)(4p 2 9) 2(8κ) 2 (4p2 1)(4p 2 9) 2!(8κ) 2 Replace (4p2 1)(4p 2 9)(4p 2 25) 2(8κ) 3 (4p2 1)(4p 2 9)(4p 2 25) 3!(8κ) 3 350 10 = (A10) Replace the integral in the numerator 1 1 teκt (1 t 2 ) (p 3)/2 dt 398 19 Replace reconstruction reconstructions 398 14 Insert matrix before Langevin 398 13 Replace Matsuita Matsusita 398 5 Replace likelhood likelihood 398 2 Replace 53 54 402 11 Replace 1980 1981 404 26 Replace (1999) (2001) 404 27 Replace Submitted for publication J Multivariate Anal, 77, 1 20 406 3 Insert matrix before distribution 408 16 Replace A-s 86 AS 86 I am very grateful to all those who have pointed out errors in the text PEJ 21/2/14 11