corrplot plotting on strange x and y axes values

I have a dataset (Metdata, attached) that I want to use corrplot (attached) with.This works but the X axis and Y is much larger for the scatter plots. It seems that it gets stuck on the histogram x and y axes. For example the visibility plots x-axis go over 2000, but visibility values < 500. Is there a way to change the axes values without zooming in independently on each plot?

Matlab version: Matlab R2013a

[Ryan]

I am having the exact same problem. I have a 12 x 12 correlation plot. All of which have data points compressed to the left so that you cannot see any trend at a glance. The data range in some plots only goes up to 500, yet the axis extends to 3000. Did you find a solution to this?

Answer 1

[Adam Danz]

Yeah, this is annoying.

You can loop through each scatter plot and set the x and y limits to the x,y ranges within the data. Since the x and y data ranges are the same within rows and within columns, you'll still have identical axis limits within rows and columns. This example uses the "MetData" dataset proivided by the OP.

[~, ~, h] = corrplot(MetData);                         %grab handle to plot objects
lineHandles = h(strcmp(get(h, 'type'), 'line'));       %get handles for scatter plots only

% Loop through each scatter plot
for i = 1:numel(lineHandles)
    x = lineHandles(i).XData;                         %x data 
    y = lineHandles(i).YData;                         %y data
    xlim(lineHandles(i).Parent, [min(x), max(x)]);    % set x limit to range of x data
    ylim(lineHandles(i).Parent, [min(y), max(y)]);    % set y limit to range of y data
    
    % To convince yourself that the axis scales are still the same within rows/cols,
    % include these two lines of code that will display tick marks.
    %lineHandles(i).Parent.Position(3:4) = lineHandles(i).Parent.Position(3:4) * .8; 
    %set(lineHandles(i).Parent, 'XTickMode', 'auto', 'XTickLabelMode', 'auto', 'YTickMode', 'auto', 'YTickLabelMode', 'auto')
end

% now take care of the x axis limits of the histogram plots
histHandles = h(strcmp(get(h, 'type'), 'histogram'));     %handles to all hist plots
% loop through hist plots
for j = 1:numel(histHandles)
    x = histHandles(j).BinEdges;                         %bin edges
    xlim(histHandles(j).Parent, [min(x), max(x)]);       %set x limits
end

Not sure why the last histogram has tiny bins (that's not from my code).

[Ryan]

That works perfectly, thanks so much.

Unfortunately I am not the original poster, but this definitely solved my problem.

[Adam Danz]

Thanks for the confirmation. It helps future visitors to know that this solution worked for the given problem.

[Carlos Borau]

This worked but at least in my case I needed to tweak manually the x limits to adjust the x labels at the bottom.

In the histHandle loop I added:

curr_fig.Children(nvars*nvars+(nvars+1)).UserData{1}(nvars,j).XLim = histhandles(j).BinLimits;

where

nvars = numel(histHandles)

and

curr_fig = gcf 

(be aware, not the handle returned by the corrplot function!). That UserData field stores several axes, some of them repeated. Maybe there is a more elegant way to do it, but I could not find it.

[Adam Danz]

Thanks for your comment, Carlos. Why did you need to adjust the x-axis limits? The histogram bins should have different axis limits than the scatter plots. For example, the x-axis limits for the 4x4 axes in my answer are shown below. Each column of axes contain the same x-axis limits except along the diagonal which contains the histograms (same with the rows of the y-axis limits, not shown). The xlim of historgrams should include the xlim of the scatter axes but they don't have to be the same.

[ 0 510] [0 14.3] [0.7 21.5] [0 3.8]

[1.3 500] [0 14.5] [0.7 21.5] [0 3.8]

[1.3 500] [0 14.3] [0.5 21.5] [0 3.8]

[1.3 500] [0 14.3] [0.7 21.5] [0 3.8]

To see the xlims for your plot,

[~, ~, h] = corrplot(___);  
ax = arrayfun(@(h)ancestor(h,'Axes'), h);  % axis handles
reshape(get(ax, 'xlim'),size(h)) % xlim array