Skip to content
Snippets Groups Projects
Commit 936ca6fd authored by JuanPi Carbajal's avatar JuanPi Carbajal
Browse files

add visualization of bivariate normal distribution

parent 32906459
No related branches found
No related tags found
No related merge requests found
......@@ -33,7 +33,7 @@ Session 1.1: Overview. Filtering and smoothing.
Session 1.2: Iterated maps. Ordinary differential equations (ODEs). Discretization of ODEs.
Session 1.3: Error propagation. Gaussain distribution.
Session 1.3: Error propagation. Gaussian distribution.
Session 1.4: Stochastic modelling. Statistical dependence and causality.
......
pkg load statistics
%This script illustrates a multivariate Gaussian distribution and its
%marginal distributions
%2-d Mean and covariance matrix
vx = 1; vy = 1; vxy = -0.5;
MeanVec = [0 0];
CovMatrix = [vx vxy; vxy vy];
%Get the sigma ellipses by transform a circle by the cholesky decomp
L = chol(CovMatrix,'lower');
t = linspace(0,2*pi,100); %Our ellipse will have 100 points on it
C = [cos(t) ; sin(t)]; %A unit circle
E1 = 1*L*C; E2 = 2*L*C; E3 = 3*L*C; %Get the 1,2, and 3-sigma ellipses
%Define limits of plotting
s_max = sqrt(max(sumsq(5*L*C,1)))
X = linspace(-s_max, s_max, 30);
Y = linspace(-s_max, s_max, 30);
%Get the 1-d PDFs for the "walls"
Z_x = normpdf(X,MeanVec(1), sqrt(CovMatrix(1,1)));
Z_y = normpdf(Y,MeanVec(2), sqrt(CovMatrix(2,2)));
%Get the 2-d samples for the "floor"
Samples = mvnrnd(MeanVec, CovMatrix, 10000);
figure; hold on;
%Plot the samples on the "floor"
plot3(Samples(:,1),Samples(:,2),zeros(size(Samples,1),1),'k.','MarkerSize',2)
%Plot the 1,2, and 3-sigma ellipses slightly above the floor
%plot3(E1(1,:), E1(2,:), 1e-3+zeros(1,size(E1,2)),'Color','g','LineWidth',2);
%plot3(E2(1,:), E2(2,:), 1e-3+zeros(1,size(E2,2)),'Color','g','LineWidth',2);
%plot3(E3(1,:), E3(2,:), 1e-3+zeros(1,size(E3,2)),'Color','g','LineWidth',2);
%Plot the histograms on the walls from the data in the middle
%[n_x, xout] = hist(Samples(:,1),20);%Creates 20 bars
%n_x = n_x ./ ( sum(n_x) *(xout(2)-xout(1)));%Normalizes to be a pdf
%[~,~,~,x_Pos,x_Height] = makebars(xout,n_x);%Creates the bar points
%plot3(x_Pos, Y(end)*ones(size(x_Pos)),x_Height,'-k')
%Now plot the other histograms on the wall
%[n_y, yout] = hist(Samples(:,2),20);
%n_y = n_y ./ ( sum(n_y) *(yout(2)-yout(1)));
%[~,~,~,y_Pos,y_Height] = makebars(yout,n_y);
%plot3(X(1)*ones(size(y_Pos)),y_Pos, y_Height,'-k')
%stem3(X(1)*ones(size(yout)),yout, n_y,'-k')
%Now plot the 1-d pdfs over the histograms
plot3(X, ones(size(X))*Y(end), Z_x,'-b','LineWidth',2);
plot3(ones(size(Y))*X(1), Y, Z_y,'-r','LineWidth',2);
# Plot the pdf surface
[xx, yy] = meshgrid(X, Y);
rr = [xx(:) yy(:)];
dr = rr - MeanVec;
iCov = cholinv(CovMatrix);
zz_pdf = exp(-0.5 * sum((dr * iCov) .* dr, 2)) / 2 / pi / sqrt(det(CovMatrix));
zz = reshape(zz_pdf, size(xx));
contour3(xx, yy, zz, 25, 'g');
%Make the figure look nice
grid on; view(45,55);
axis([X(1) X(end) Y(1) Y(end)])
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment