Multivariate Statistics Fundamentals Part 1: Rotation-based Techniques

Similar documents
Linear Algebra: Matrix Eigenvalue Problems

Principle Components Analysis (PCA) Relationship Between a Linear Combination of Variables and Axes Rotation for PCA

Linear Algebra Section 2.6 : LU Decomposition Section 2.7 : Permutations and transposes Wednesday, February 13th Math 301 Week #4

Data Preprocessing Tasks

Multivariate Data Analysis a survey of data reduction and data association techniques: Principal Components Analysis

CS 246 Review of Linear Algebra 01/17/19

Mathematical foundations - linear algebra

Linear Algebra. Matrices Operations. Consider, for example, a system of equations such as x + 2y z + 4w = 0, 3x 4y + 2z 6w = 0, x 3y 2z + w = 0.

PCA, Kernel PCA, ICA

Vectors and Matrices Statistics with Vectors and Matrices

I L L I N O I S UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN

Econ Slides from Lecture 7

1. Introduction to Multivariate Analysis

Lecture 2: Linear Algebra Review

JUST THE MATHS SLIDES NUMBER 9.6. MATRICES 6 (Eigenvalues and eigenvectors) A.J.Hobson

Linear Algebra review Powers of a diagonalizable matrix Spectral decomposition

Elementary Row Operations on Matrices

Eigenvalues, Eigenvectors, and an Intro to PCA

Review of Basic Concepts in Linear Algebra

Math Spring 2011 Final Exam

Knowledge Discovery and Data Mining 1 (VO) ( )

Principal Component Analysis

Math for ML: review. CS 1675 Introduction to ML. Administration. Lecture 2. Milos Hauskrecht 5329 Sennott Square, x4-8845

Linear Algebra review Powers of a diagonalizable matrix Spectral decomposition

M.A.P. Matrix Algebra Procedures. by Mary Donovan, Adrienne Copeland, & Patrick Curry

Linear Algebra Review. Vectors

The Singular Value Decomposition (SVD) and Principal Component Analysis (PCA)

2. Matrix Algebra and Random Vectors

ICS 6N Computational Linear Algebra Symmetric Matrices and Orthogonal Diagonalization

Properties of Linear Transformations from R n to R m

Principal Components Analysis (PCA)

AN ITERATION. In part as motivation, we consider an iteration method for solving a system of linear equations which has the form x Ax = b

Determinants by Cofactor Expansion (III)

STATISTICS 407 METHODS OF MULTIVARIATE ANALYSIS TOPICS

Introduction to Matrices

10. Linear Systems of ODEs, Matrix multiplication, superposition principle (parts of sections )

Appendix A: Matrices

Cheng Soon Ong & Christian Walder. Canberra February June 2017

Chapters 5 & 6: Theory Review: Solutions Math 308 F Spring 2015

Eigenvalues, Eigenvectors, and an Intro to PCA

Eigenvalues, Eigenvectors, and an Intro to PCA

Math for ML: review. ML and knowledge of other fields

Unsupervised Learning: Dimensionality Reduction

Principal Component Analysis!! Lecture 11!

Linear Algebra Practice Problems

Basic Concepts in Matrix Algebra

Principal Component Analysis (PCA) Theory, Practice, and Examples

Chapter 3. Linear and Nonlinear Systems

Basic Concepts in Linear Algebra

Finite Math - J-term Section Systems of Linear Equations in Two Variables Example 1. Solve the system

Principal component analysis

Linear Algebra & Geometry why is linear algebra useful in computer vision?

Focus was on solving matrix inversion problems Now we look at other properties of matrices Useful when A represents a transformations.

Linear & Non-Linear Discriminant Analysis! Hugh R. Wilson

Linear Algebra Solutions 1

Linear Algebra & Geometry why is linear algebra useful in computer vision?

Multivariate Statistical Analysis

Principal Component Analysis

Matrices: 2.1 Operations with Matrices

I L L I N O I S UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN

7.5 Operations with Matrices. Copyright Cengage Learning. All rights reserved.

Linear vector spaces and subspaces.

MATH 423 Linear Algebra II Lecture 33: Diagonalization of normal operators.

Mathematical foundations - linear algebra

Linear Algebra (Review) Volker Tresp 2018

Factor Analysis and Kalman Filtering (11/2/04)

1. Linear systems of equations. Chapters 7-8: Linear Algebra. Solution(s) of a linear system of equations (continued)

Singular Value Decomposition and Principal Component Analysis (PCA) I

Prepared by: M. S. KumarSwamy, TGT(Maths) Page

Dimensionality Reduction: PCA. Nicholas Ruozzi University of Texas at Dallas

Numerical Linear Algebra Homework Assignment - Week 2

MATH Topics in Applied Mathematics Lecture 12: Evaluation of determinants. Cross product.

Math Linear Algebra Final Exam Review Sheet

Homework 1 Elena Davidson (B) (C) (D) (E) (F) (G) (H) (I)

Multilevel Models in Matrix Form. Lecture 7 July 27, 2011 Advanced Multivariate Statistical Methods ICPSR Summer Session #2

MATH 829: Introduction to Data Mining and Analysis Principal component analysis

6 EIGENVALUES AND EIGENVECTORS

1 Multiply Eq. E i by λ 0: (λe i ) (E i ) 2 Multiply Eq. E j by λ and add to Eq. E i : (E i + λe j ) (E i )

A Introduction to Matrix Algebra and the Multivariate Normal Distribution

A Tutorial on Data Reduction. Principal Component Analysis Theoretical Discussion. By Shireen Elhabian and Aly Farag

Math Matrix Algebra

Recall : Eigenvalues and Eigenvectors

CS 143 Linear Algebra Review

Linear Algebra: Characteristic Value Problem

Chapter 6. Eigenvalues. Josef Leydold Mathematical Methods WS 2018/19 6 Eigenvalues 1 / 45

More Linear Algebra. Edps/Soc 584, Psych 594. Carolyn J. Anderson

Gershgorin s Circle Theorem for Estimating the Eigenvalues of a Matrix with Known Error Bounds

3 (Maths) Linear Algebra

Chapter 3. Determinants and Eigenvalues

Math 315: Linear Algebra Solutions to Assignment 7

JUST THE MATHS UNIT NUMBER 9.9. MATRICES 9 (Modal & spectral matrices) A.J.Hobson

Principal Component Analysis -- PCA (also called Karhunen-Loeve transformation)

Math 489AB Exercises for Chapter 2 Fall Section 2.3

Dimensionality Reduction and Principal Components

Lecture: Face Recognition and Feature Reduction

Elementary maths for GMT

Chapter 2. Linear Algebra. rather simple and learning them will eventually allow us to explain the strange results of

Chapter 17: Undirected Graphical Models

Math Camp Notes: Linear Algebra II

j=1 u 1jv 1j. 1/ 2 Lemma 1. An orthogonal set of vectors must be linearly independent.

Transcription:

Multivariate Statistics Fundamentals Part 1: Rotation-based Techniques

A reminded from a univariate statistics courses Population Class of things (What you want to learn about) Sample group representing a class (What you actually study) Experimental unit individual research subject (e.g. location, entity, etc.) Response variable property of thing that you believe is the result of predictors (What you actually measure e.g. tree height) a.k.a dependent variable Predictor variable(s) environment of things which you believe is influencing a response variable (e.g. climate, soil attributes, topography, etc.) a.k.a independent variable Error - difference between an observed value (or calculated) value and its true (or expected) value

Experimental Unit (row) Data.I D 1 2 Type Varable 1 2 3 4 3 4 5 6 Regions Ecosystems Forest types Treatments, etc. In Multivariate statistics: Frequency of species Climate variables Soil characteristics Nutrient concentrations Contaminants, etc. s can be either numeric or categorical (depends on the technique) Focus is often placed on graphical representation of results

3 3, 6, 8 2 Rotation-based techniques Data.I D Type Varable 1 2 3 4 1 2 3 4 5 6 1 Find an equation to rotate data to so that axis explains multiple variables Final results based on multiple variables give different inferences than 2 variables Repeat rotation process to achieve analysis objective 1,2 1, 2, 4, 9, 10

Objectives of Rotation-based techniques 1. Rotate so that new axis explains the greatest amount of variation within the data Principal Component Analysis (PCA) Factor Analysis 2. Rotate so that the variation between groups is maximized Multivariate Analysis of Variance (MANOVA) Discriminant Analysis (DISCRIM) 3. Rotate so that one dataset explains the most variation in another dataset Canonical Correspondence Analysis (CCA)

Component 2 (y ) Math behind Rotation-based techniques Trigonometry y y x = 30 Original point (x,y) = (1,0) y = sin(α) = 0.5 x x = cos(α) = 0.8 Point after rotation (x,y ) = (0.8,0.5) Simple to understand But difficult/time consuming to do for more than 2 variables Component 1 (x )

Matrix Algebra Allows us to simultaneously rotate the data based on MANY variables Zero Matrix 0 = 0 0 0 A = (m x n matrix) a 11 a 12 a 1n a 21 a 22 a 2n 0 0 0 0 0 0 a m1 a m2 a mn If m=n than A is a square matrix If n=1 than A is a column vector If m=1 than A is a row vector Diagonal Matrix D = d 1 0 0 0 d 2 0 0 0 d n Transpose Matrix A = a 11 a 21 a m1 Identity Matrix I = 1 0 0 a 12 a 22 a m2 0 1 0 a 1n a 2n a nm 0 0 1

Matrix Algebra Allows us to simultaneously rotate the data based on MANY variables A + B = a 11 a 12 a 1n a 21 a 22 a 2n + b 11 b 12 b 1n b 21 b 22 b 2n = a 11 +b 11 a 12 +b 12 a 1n +b 1n a 21 +b 21 a 22 +b 22 a 2n +b 2n a m1 a m2 a mn b m1 b m2 b mn a m1 +b m1 a m2 +b m2 a mn +b mn A - B = a 11 a 12 a 1n a 21 a 22 a 2n - b 11 b 12 b 1n b 21 b 22 b 2n = a 11 -b 11 a 12 -b 12 a 1n -b 1n a 21 -b 21 a 22 -b 22 a 2n -b 2n a m1 a m2 a mn b m1 b m2 b mn a m1 -b m1 a m2 -b m2 a mn -b mn Addition & subtraction only works if matrices A and B are the same size

Matrix Algebra Allows us to simultaneously rotate the data based on MANY variables When multiplying matrices the number of columns in matrix A has to be equal to the number of rows in matrix B A * B = a 11 a 12 a 1n a 21 a 22 a 2n * b 11 b 12 b 1p b 21 b 22 b 2p = Ʃa 1j b j1 Ʃa 1j b j2 Ʃa 1j b jp Ʃa 2j b j1 Ʃa 2j b j2 Ʃa 2j b jp a m1 a m2 a mn b m1 b m2 b np Ʃa mj b j1 Ʃa mj b j2 Ʃa mj b jp Where j is the column number in A and the row number in B AND: Ʃ a ij b jk = a i1 b 1k + a i2 b 2k + + a in b nk You multiple the cells across from A and down from B The final output is a column of values of length n (number of columns in A and rows in B)

Matrix Algebra Allows us to simultaneously rotate the data based on MANY variables k*a = ka 11 ka 12 ka 1n ka 21 ka 22 ka 2n Multiplying by a scaler (coefficient) ka m1 ka m2 ka mn Calculating an inverse more than an 2D matrix is tedious to do by hand R will do it for you M -1 = a b c d = d/ -b/ -c/ a/ Where = (a * d)- (b * c)

Eigenvalues & Eigenvectors Consider the set of equations: a 11 x 1 + a 12 x 2 + + a 1n x n = λx 1 a 21 x 1 + a 22 x 2 + + a 2n x n = λx 2 a n1 x 1 + a n2 x 2 + + a nn x n = λx n Can also be written in matrix form as: Ax = λx or (A-λI)x = 0 where I is the n x n identity matrix And ) is column vector of 0 s Equations will only hold true for particular values of λ each value is an eigenvalue (up to n values) Solving the equations for a particular eigenvalue will require a set of values for x, where x = - this is an eigenvector x 1 X 2 x n