Inleiding Multivariate Statistiek Assignment 1 Exercise 1 This exercise is made with R. The code which is used to make the exercise is appended. 1. Table with distances: (Note: the column Mahalanobis Distance² is (x – u)T * S^-1 * (x – u), while the column Mahalanobis Distance is the square root of (x – u)T * S^-1 * (x – u)). State: -> Mahalanobis Mahalanobis State: -> Mahalanobis Mahalanobis Distance²: Distance: Distance²: Distance: "Alabama" 1,8263 1,3514 "Montana" 2,8459 1,6870 "Alaska" 12,0269 3,4680 "Nebraska" 1,6850 1,2981 "Arizona" 0,0781 0,2796 "Nevada" 0,7125 0,8441 "Arkansas" 1,1622 1,0780 "New Hampshire" 0,3326 0,5767 "California" 6,4517 2,5400 "New Jersey" 4,1053 2,0262 "Colorado" 0,2178 0,4667 "New Mexico" 2,0500 1,4318 "Connecticut" 7,5033 2,7392 "New York" 4,6069 2,1464 "Delaware" 1,9284 1,3887 "North Carolina" 4,3855 2,0941 "Florida" 2,6585 1,6305 "North Dakota" 8,2045 2,8643 "Georgia" 4,1647 2,0408 "Ohio" 0,2029 0,4504 "Hawaii" 12,1474 3,4853 "Oklahoma" 2,4395 1,5619 "Idaho" 1,5440 1,2426 "Oregon" 0,3589 0,5990 "Illinois" 4,5440 2,1317 "Pennsylvania" 3,4479 1,8569 "Indiana" 2,1426 1,4638 "Rhode Island" 2,6340 1,6230 "Iowa" 3,6021 1,8979 "South Carolina" 6,1565 2,4812 "Kansas" 2,3067 1,5188 "South Dakota" 1,9659 1,4021 "Kentucky" 1,1529 1,0737 "Tennessee" 2,5316 1,5911 "Louisiana" 2,3279 1,5257 "Texas" 48,0200 6,9296 "Maine" 2,1632 1,4708 "Utah" 3,2598 1,8055 "Maryland" 1,0750 1,0368 "Vermont" 1,0874 1,0428 "Massachusetts" 1,1969 1,0940 "Virginia" 1,7161 1,3100 "Michigan" 3,9363 1,9840 "Washington" 0,2264 0,4758 "Minnesota" 4,8073 2,1926 "West Virginia" 2,4765 1,5737 "Mississippi" 3,3095 1,8192 "Wisconsin" 4,4089 2,0997 "Missouri" 1,2072 1,0987 "Wyoming" 0,6586 0,8115 2. Table with angles: Deviation vector 1: Salary Salary Salary Verbal Verbal Math Deviation vector 2; Verbal Math Total Math Total Total Angle (in degrees): 118,749° 114,229° 116,580° 14,559° 7,583° 7,127° 3. The calculated angle is approximately: 0,083°. Note: expected was 0, because the vectors TOTAL and VERBAL + MATH are the same by definition. However, in the file the VERBAL score and MATH score from Texas do not add up to the TOTAL score. This is why there is a small angle between the vectors. Exercise 2 1. Part 1: 2. Part 2: For clarification of the axis: X AND Y Y AND Z X AND Z PROB. MASS 90% 99% 90% 99% 90% 99% AXIS 1 (1.86, 1.86) (2.63, 2.63) (2.15, 0) (3.03, 0) (2.07, 2.07) (2.93, 2.93) AXIS 2 (-1.07, 1.07) (-1.52, 1.52) (0, 2.15) (0, 3.03) (0.57, 0.57) (0.79, -0.79) 3. The change of the expected value will not change the shape of the ellipses from 2. The only thing that will change is around what point the ellipses will be centered. Before the change the ellipses were centered around (0,0,0). After the change the ellipses are going to be centered around (3,3,3), but they are still going to have the same shape as before. Appendix Code for exercise 1: R code: dataset = read.table(file.choose(), header = TRUE, row.names = 1) # Creating the dataset by choosing the correct file. MahalanobisDistance = mahalanobis(dataset, colMeans(dataset), cov(dataset)) # Calculating the Mahalanobis distance of the sample. MahalanobisDistanceTable = data.frame(MahalanobisDistance) # Making a table of the Mahalanobis distance. A = colMeans(dataset) # Computing the mean for each variable and storing the results. SalaryMean = A[1] VerbalMean = A[2] MathMean = A[3] TotalMean = A[4] DeviationVectorSalary = dataset[,1] - SalaryMean # Computing the deviation vector for each variable. DeviationVectorVerbal = dataset[,2] - VerbalMean DeviationVectorMath = dataset[,3] - MathMean DeviationVectorTotal = dataset[,4] - TotalMean # Defining function to calculate the angle between two vectors. CalculateAngle = function(a,b){ return (180 * acos(sum(a*b) / ( sqrt(sum(a^2)) * sqrt(sum(b^2)))) / pi) } AngleSalaryVerbalDeviation = CalculateAngle(DeviationVectorSalary, DeviationVectorVerbal) # Calculating all the angles. AngleSalaryMathDeviation = CalculateAngle(DeviationVectorSalary, DeviationVectorMath) AngleSalaryTotalDeviation = CalculateAngle(DeviationVectorSalary, DeviationVectorTotal) AngleVerbalMathDeviation = CalculateAngle(DeviationVectorVerbal, DeviationVectorMath) AngleVerbalTotalDeviation = CalculateAngle(DeviationVectorVerbal, DeviationVectorTotal) AngleMathTotalDeviation = CalculateAngle(DeviationVectorMath, DeviationVectorTotal) AngleTotalAndVerbalPlusMath = CalculateAngle(dataset[,4], dataset[,2] + dataset[,3]) # Calculating the angle between TOTAL and VERBAL + MATH.