电脑实习课讲义(二).docVIP

PAGE 1 PAGE 5 中華大學應用數學系 應用多變量分析 電腦實習課講義(二) 羅 琪 老師編 內容包含: The Multivariate Normal Checking Multivariate Normality The Multivariate Normal We look at only the bivariate normal density because we can plot it. First we need a function for computing the bivariate normal density. , , dbinorm_function(x,mu,sigma) + {det_sigma[1,1]*sigma[2,2]-sigma[1,2]*sigma[2,1] + a_t(x-mu) %*% solve(sigma) %*% (x-mu) + exp(-a/2)/(sqrt(det)*2*pi) + } dbinorm function(x, mu, sigma) { det - sigma[1, 1] * sigma[2, 2] - sigma[1, 2] * sigma[2, 1] a - t(x - mu) %*% solve( sigma) %*% (x - mu) exp( - a/2)/(sqrt(det) * 2 * pi) } Now create a zero mean vector and an identity variance matrix: mu_c(0,0) # mu [1] 0 0 sigma_matrix(c(1,0,0,1),ncol=2) # ncol = # of columns 行數, sigma [,1] [,2] [1,] 1 0 [2,] 0 1 Now create the range of values for which we wish to plot the density: x_seq(-3,3,length=20) # Creates a vector of evenly spaced numbers. Length-number of values in the result. y_seq(-3,3,length=20) y [1] -3.0000000 -2.6842105 -2.3684211 [4] -2.0526316 -1.7368421 -1.4210526 [7] -1.1052632 -0.7894737 -0.4736842 [10] -0.1578947 0.1578947 0.4736842 [13] 0.7894737 1.1052632 1.4210526 [16] 1.7368421 2.0526316 2.3684211 2.6842105 3.0000000 A matrix for storing the density at point (x[I], y[j]): z_matrix(0,ncol=20,nrow=20) # nrow = # of rows 列數 Compute the density over the matrix: for(i in 1:20){ + for(j in 1:20){ + z[i,j]_dbinorm(c(x[i],y[j]),mu,sigma) + }} Plot the result in two ways: persp(x,y,z) # Creates a perspective plot 雙變量常態分配密度函數的透視圖 contour(x,y,z) # Make a contour plot Now let’s construct a distribution with some correlation. I choose this to have determinant one so that the scale will be similar to the density case. What is the population correlation coefficient here? sigmac_matrix(c(sqrt(2),1,1,sqrt(2)),ncol=2) # , sigma