undefined function 'corrplot

Hi, I wanted to create a graphical correlation matrix according using the corrplot function ( http://www.mathworks.com/help/toolbox/econ/corrplot.html) but when I try using the function I get the following message:

Undefined function 'corrplot' for input arguments of type 'double'.

I also looked for the function using 'which'- command. And I couldn't find the functiin. Is it possible that my Matlab version (2011b) doesn't have the function? Is it possible to download the function on the Mathworks website?

Thank you.

Answer 1

[Adam Danz]

corrplot is part of the Econometrics Toolbox. Run this command to determine if you have the Econometrics Toolbox installed on your machine. It returns true (1) or false (0).

license('test','econometrics_toolbox')

Alternative

plotmatrix() is quite similar and with just a few additional lines, you can recreate most of the corrplot().

Here's a demo.

% Load data
load('carsmall')

% Unlike corrplot, plotmatrix cannot receive a table. 
% Form a matrix of variables in each column.
data = [Horsepower, Weight, Acceleration, MPG]; 
nVars = size(data,2); 

% Remove any rows that contain NaN values. Otherwise corr() will 
% return NaN. 
data(any(isnan(data),2), :) = []; 

% Create plotmatrix
figure('Name', 'carsmall_data')
[sh, ax, ~, hh] = plotmatrix(data);

% Add axis labels
arrayfun(@(h,lab)ylabel(h,lab),ax(:,1), {'Horsepower','Weight','Accel.','MPG'}')
arrayfun(@(h,lab)xlabel(h,lab),ax(end,:), {'Horsepower','Weight','Accel.','MPG'})

% Compute correlation for each scatter plot axis
[r,p] = arrayfun(@(h)corr(h.Children.XData(:),h.Children.YData(:)),ax(~eye(nVars)));

% Label the correlation and p values
arrayfun(@(h,r,p)text(h,min(xlim(h))+range(xlim(h))*.05,max(ylim(h)),...
    sprintf('r=%.2f,  p=%.3f',r,p),'Horiz','Left','Vert','top','FontSize',8,'Color','r'),...
    ax(~eye(nVars)),r,p)

% Change marker appearance
set(sh, 'Marker', 'o','MarkerSize', 2, 'MarkerEdgeColor', ax(1).ColorOrder(1,:)) 
lsh = arrayfun(@(h)lsline(h),ax(~eye(nVars)));

% Add least square regression line. 
set(lsh,'color', 'm')

[Adam Danz]

Note: based on conversation with dpb, the correlation coefficient and the slope of the regression line are equal after standardizing the data. For example (comparing Weight and Horsepower)

load('carsmall')
rho = corr([Weight, Horsepower],'type','pearson','rows','pairwise','tail','both')

rho = 2×2

1.0000 0.8733 0.8733 1.0000

n = isnan(Weight) | isnan(Horsepower); 
polyfit(zscore(Weight(~n)),zscore(Horsepower(~n)),1)

ans = 1×2

0.8733 0.0000

[ClaireB]

Thanks for the nice example. i would like to use Spearman correlation instead of pearson but unsure where to add this in the following function

[r,p] = arrayfun(@(h)corr(h.Children.XData(:),h.Children.YData(:)),ax(~eye(nVars)));

[Adam Danz]

Set the correlation type using a name-value pair

[r,p] = arrayfun(@(h)corr(h.Children.XData(:),h.Children.YData(:),'Type','Spearman'),ax(~eye(nVars)));

Answer 2

[owr]

It looks like it was added in 2012A. I dont think there is anywhere to download just one function, too many dependencies, best thing would be to install 2012A if possible.

Answer 3

[Anni K]

Hi, I had the same problem. After downloading the toolbox Econometric Modeler, corrplot worked for me.

Good luck!