Presentation is loading. Please wait.

Presentation is loading. Please wait.

NOTES ON MULTIPLE REGRESSION USING MATRICES  Multiple Regression Tony E. Smith ESE 502: Spatial Data Analysis  Matrix Formulation of Regression  Applications.

Similar presentations


Presentation on theme: "NOTES ON MULTIPLE REGRESSION USING MATRICES  Multiple Regression Tony E. Smith ESE 502: Spatial Data Analysis  Matrix Formulation of Regression  Applications."— Presentation transcript:

1 NOTES ON MULTIPLE REGRESSION USING MATRICES  Multiple Regression Tony E. Smith ESE 502: Spatial Data Analysis  Matrix Formulation of Regression  Applications to Regression Analysis

2 SIMPLE LINEAR MODEL  Data:  Parameters:  Model:

3 SIMPLE REGRESSION ESTIMATION  Data Points:  Predicted Value:  Estimate Conditional Mean:    Line of Best Fit where:

4 STANDARD LINEAR MODEL  Data:  Parameters:  Model:

5 STANDARD LINEAR MODEL (k = 2)  Data:  Parameters:  Model:

6 REGRESSION ESTIMATION (for k =2) Plane of Best Fit     Data Points:  Predicted Value: where:

7 MATRIX REPRESENTATION OF THE STANDARD LINEAR MODEL  Vectors and Matrices:  Matrix Reformulation of the Model:

8 LINEAR TRANSFORMATIONS IN ONE DIMENSION  Linear Function:  Graphic Depiction:   

9 LINEAR TRANSFORMATIONS IN TWO DIMENSIONS  Linear Transformation:

10  Graphical Depiction of Linear Transformation:        

11 SOME MATRIX CONVENTIONS  Transposes of Vectors and Matrices:  Symmetric (Square) Matrices:  Important Example:

12  Column Representation of Matrices:  Row Representation of Matrices:

13  Matrix Multiplication:  Inner Product of Vectors:  Transposes:

14 MATRIX REPRESENTATIONS OF LINEAR TRANSFORMATIONS  For any Two-Dimensional Linear Transformation : with :

15  Graphical Depiction of Matrix Representation:        

16  Inversion of Square Matrices (as Linear Transformations):

17 DETERMINANTS OF SQUARE MATRICES      

18 NONSINGULAR SQUARE MATRICES      

19 LEAST-SQUARES ESTIMATION  General Sum-of-Squares:  General Regression Matrices:

20 DIFFERENTIATION OF FUNCTIONS  General Derivative:  Example:   

21 PARTIAL DERIVATIVES  

22 VECTOR DERIVATIVES  Derivative Notation for:  Gradient Vector:

23 TWO IMPORTANT EXAMPLES  Linear Functions:  Quadratic Functions:

24  Quadratic Derivatives:  Symmetric Case:

25 MINIMIZATION OF FUNCTIONS  First-Order Condition:  Example:  

26    TWO-DIMENSIONAL MINIMIZATION

27 LEAST SQUARES ESTIMATION  Solution for:

28 NON-MATRIX VERSION (k = 2)  Data:  Beta Estimates:

29 EXPECTED VALUES OF RANDOM MATRICES  Random Vectors and Matrices  Expected Values:

30 EXPECTATIONS OF LINEAR FUNCTIONS OF RANDOM VECTORS  Linear Combinations  Linear Transformations

31 EXPECTATIONS OF LINEAR FUNCTIONS OF RANDOM MATRICES  Left Multiplication  Right Multiplication (by symmetry of inner products):

32 COVARIANCE OF RANDOM VECTORS  Random Variables :  Random Vectors:

33 COVARIANCE OF LINEAR FUNCTIONS OF RANDOM VECTORS  Linear Combinations:  Linear Transformations: ( Right Mult ) ( Left Mult )

34 TRANSLATIONS OF RANDOM VECTORS  Translation:  Means:  Covariances:

35 RESIDUAL VECTOR IN THE STANDARD LINEAR MODEL  Linear Model Assumption:  Residual Means:  Residual Covariances:

36 MOMENTS OF BETA ESTIMATES  Linear Model:  Mean of Beta Estimates: (Unbiased Estimator)  Covariance of Beta Estimates:

37 ESTIMATION OF RESIDUAL VARIANCE  Residual Variance :  Residual Estimates :  Natural Estimate of Variance :  Bias-Correct Estimate of Variance : (Compensates for Least Squares)

38 ESTIMATION OF BETA COVARIANCE  Beta Covariance Matrix:  Beta Covariance Estimates:


Download ppt "NOTES ON MULTIPLE REGRESSION USING MATRICES  Multiple Regression Tony E. Smith ESE 502: Spatial Data Analysis  Matrix Formulation of Regression  Applications."

Similar presentations


Ads by Google