1. 1 Multivariate Linear Regression Models Shyh-Kang Jeng Department of Electrical Engineering/ Graduate Institute of Communication/ Graduate Institute of Networking and Multimedia

2. 2 Regression Analysis A statistical methodology –For predicting value of one or more response (dependent) variables –Predict from a collection of predictor (independent) variable values

3. 3 Example 7.1 Fitting a Straight Line Observed data Linear regression model z1z1z1z101234 y14389

4. 4 Example 7.1 Fitting a Straight Line 032145 z y 2 4 6 8 10 0

5. 5 Classical Linear Regression Model

6. 6

7. 7 Example 7.1

8. 8 Examples 6.6 & 6.7

9. 9 Example 7.2 One-Way ANOVA

10. 10 Method of Least Squares

11. 11 Result 7.1

12. 12 Proof of Result 7.1

13. 13 Proof of Result 7.1

14. 14 Example 7.1 Fitting a Straight Line Observed data Linear regression model z1z1z1z101234 y14389

15. 15 Example 7.3

16. 16 Coefficient of Determination

17. 17 Geometry of Least Squares

18. 18 Geometry of Least Squares

19. 19 Projection Matrix

20. 20 Result 7.2

21. 21 Proof of Result 7.2

22. 22 Proof of Result 7.2

23. 23 Result 7.3 Gauss Least Square Theorem

24. 24 Proof of Result 7.3

25. 25 Result 7.4

26. 26 Proof of Result 7.4

27. 27 Proof of Result 7.4

28. 28 Proof of Result 4.11

29. 29 Proof of Result 7.4

30. 30 Proof of Result 7.4

31. 31 Proof of Result 7.4

32. 32  2 Distribution

33. 33 Result 7.5

34. 34 Proof of Result 7.5

35. 35 Example 7.4 (Real Estate Data) 20 homes in a Milwaukee, Wisconsin, neighborhood Regression model

36. 36 Example 7.4

37. 37 Result 7.6

38. 38 Effect of Rank In situations where Z is not of full rank, rank(Z) replaces r+1 and rank(Z 1 ) replaces q+1 in Result 7.6

39. 39 Proof of Result 7.6

40. 40 Proof of Result 7.6

41. 41 Wishart Distribution

42. 42 Generalization of Result 7.6

43. 43 Example 7.5 (Service Ratings Data)

44. 44 Example 7.5: Design Matrix

45. 45 Example 7.5

46. 46 Result 7.7

47. 47 Proof of Result 7.7

48. 48 Result 7.8

49. 49 Proof of Result 7.8

50. 50 Example 7.6 (Computer Data)

51. 51 Example 7.6

52. 52 Adequacy of the Model

53. 53 Residual Plots

54. 54 Q-Q Plots and Histograms Used to detect the presence of unusual observations or severe departures from normality that may require special attention in the analysis If n is large, minor departures from normality will not greatly affect inferences about 

55. 55 Test of Independence of Time

56. 56 Example 7.7: Residual Plot

57. 57 Leverage “ Outliers ” in either the response or explanatory variables may have a considerable effect on the analysis and determine the fit Leverage for simple linear regression with one explanatory variable z

58. 58 Mallow’s C p Statistic Select variables from all possible combinations

59. 59 Usage of Mallow’s C p Statistic

60. 60 Stepwise Regression 1. The predictor variable that explains the largest significant proportion of the variation in Y is the first variable to enter 2. The next to enter is the one that makes the highest contribution to the regression sum of squares. Use Result 7.6 to determine the significance (F-test)

61. 61 Stepwise Regression 3. Once a new variable is included, the individual contributions to the regression sum of squares of the other variables already in the equation are checked using F -tests. If the F-statistic is small, the variable is deleted 4. Steps 2 and 3 are repeated until all possible additions are non- significant and all possible deletions are significant

62. 62 Treatment of Colinearity If Z is not of full rank, Z’Z does not have an inverse  Colinear Not likely to have exact colinearity Possible to have a linear combination of columns of Z that are nearly 0 Can be overcome somewhat by –Delete one of a pair of predictor variables that are strongly correlated –Relate the response Y to the principal components of the predictor variables

63. 63 Bias Caused by a Misspecified Model

64. 64 Example 7.3 Observed data Regression model z 1 z 1 0 1 2 3 4 y 1 y 1 1 4 3 8 9 y 2 y 2 -1 -1 2 3 2

65. 65 Multivariate Multiple Regression

66. 66 Multivariate Multiple Regression

67. 67 Multivariate Multiple Regression

68. 68 Multivariate Multiple Regression

69. 69 Multivariate Multiple Regression

70. 70 Multivariate Multiple Regression

71. 71 Example 7.8

72. 72 Example 7.8

73. 73 Result 7.9

74. 74 Proof of Result 7.9

75. 75 Proof of Result 7.9

76. 76 Proof of Result 7.9

77. 77 Forecast Error

78. 78 Forecast Error

79. 79 Result 7.10

80. 80 Result 7.11

81. 81 Example 7.9

82. 82 Other Multivariate Test Statistics

83. 83 Predictions from Regressions

84. 84 Predictions from Regressions

85. 85 Predictions from Regressions

86. 86 Example 7.10

87. 87 Example 7.10

88. 88 Example 7.10

89. 89 Linear Regression

90. 90 Result 7.12

91. 91 Proof of Result 7.12

92. 92 Proof of Result 7.12

93. 93 Population Multiple Correlation Coefficient

94. 94 Example 7.11

95. 95 Linear Predictors and Normality

96. 96 Result 7.13

97. 97 Proof of Result 7.13

98. 98 Invariance Property

99. 99 Example 7.12

100. 100 Example 7.12

101. 101 Prediction of Several Variables

102. 102 Result 7.14

103. 103 Example 7.13

104. 104 Example 7.13

105. 105 Partial Correlation Coefficient

106. 106 Example 7.14

107. 107 Mean Corrected Form of the Regression Model

108. 108 Mean Corrected Form of the Regression Model

109. 109 Mean Corrected Form for Multivariate Multiple Regressions

110. 110 Relating the Formulations

111. 111 Example 7.15 Example 7.6, classical linear regression model Example 7.12, joint normal distribution, best predictor as the conditional mean Both approaches yielded the same predictor of Y 1

112. 112 Remarks on Both Formulation Conceptually different Classical model –Input variables are set by experimenter –Optimal among linear predictors Conditional mean model –Predictor values are random variables observed with the response values –Optimal among all choices of predictors

113. 113 Example 7.16 Natural Gas Data

114. 114 Example 7.16 : First Model

115. 115 Example 7.16 : Second Model