代写 R C statistic DA 410 C. C HW3

DA 410 C. C HW3
Homework 3 (30 points)
1. Chapter 6 Page 245-246: #6.27 (a) (b) (c) 2. Chapter 6 Page 246-247: #6.28 (a)

DA 410
C. C
HW3
HW 3 HELP FILE
6.27 (a)
# Input the data into R
# Calculate “between” matrices
𝑯𝑯 = 𝑛𝑛 ∑ (𝒚𝒚̅𝑖𝑖. − 𝒚𝒚̅..)(𝒚𝒚̅𝑖𝑖. − 𝒚𝒚̅..) 𝑖𝑖=1
# 𝒚𝒚̅𝑖𝑖. is the column mean of method i, 𝒚𝒚̅.. is the column mean of all samples. n is the number of samples.
𝑘𝑘

method1.bar <- colMeans(method1) method2.bar <- colMeans(method2) method3.bar <- colMeans(method3)  𝒚𝒚̅ 𝑖𝑖 . method.all.bar <- (method1.bar+method2.bar+method3.bar)/3  𝒚𝒚̅ . .  𝒚𝒚̅ 𝑖𝑖 . − 𝒚𝒚̅ . . H <- 12 * unname(method1.bar.diff %*% t(method1.bar.diff) + method2.bar.diff %*% t(method2.bar.diff) + method3.bar.diff %*% t(method3.bar.diff)) method1.bar.diff <- method1.bar - method.all.bar method2.bar.diff <- method2.bar - method.all.bar method3.bar.diff <- method3.bar - method.all.bar DA 410 C. C HW3 # Calculate “within” matrices 𝑬𝑬 = ∑ ∑ (𝒚𝒚𝑖𝑖𝑗𝑗 − 𝒚𝒚̅𝑖𝑖.)(𝒚𝒚𝑖𝑖𝑗𝑗 − 𝒚𝒚̅𝑖𝑖.)′ 𝑖𝑖=1 𝑗𝑗=1 "compute.within.matrix" <- function(data, mean) { ret <- matrix(as.numeric(0), nrow=4, ncol=4) for (i in 1:12) { diff <- as.numeric(unname(data[i,] - mean)) ret <- ret + diff %*% t(diff) } return(ret) } E <- compute.within.matrix(method1, method1.bar) + compute.within.matrix(method2, method2.bar) + compute.within.matrix(method3, method3.bar) #Four MANOVA # Wilks’ Test Statistic 𝛬𝛬 = Lambda <- det(E) / det(E + H) # Pillai Statistic 𝑉𝑉(𝑠𝑠) = 𝑡𝑡𝑡𝑡[(𝐸𝐸 + 𝐻𝐻)−1𝐻𝐻] |𝑬𝑬| |𝑬𝑬+𝑯𝑯| 𝑘𝑘𝑛𝑛 Determinant of E/Determinant of E+H To calculate trace of the matrix in R, you will need to install “psych” package. > install.packages(“psych”)

DA 410 C. C HW3
> library(psych)
V.s <- tr(solve(E + H) %*% H) # Lawley-Hotelling statistic is defined as 𝑈𝑈(𝑠𝑠) = 𝑡𝑡𝑡𝑡(𝑬𝑬−𝟏𝟏𝑯𝑯) U.s <- tr(solve(E) %*% H) # Roy’s largest root test Ө = 𝜆𝜆1 𝜆𝜆 1+𝜆𝜆1 6.27 (b) See textbook page 188 example 6.1.8. 6.27 (c) Read textbook page 189-190 and see example 6.2 1 is the largest eigenvalue of matrix E lambda.1 <- eigen(solve(E) %*% H)$values[1] theta <- lambda.1 / (1 + lambda.1)