shift horizontal histogram to right

I am trying to shift the entire histogram to right, i.e. at 0.04 instead of 0, without changing the xticks.

[Ameer Hamza]

Can you show an example output. How do you want to move distogram without moving the xticks.

[Guillaume]

Personally, I find this graph very confusing. What is the benefit of showing the histogram bars horizontally rather than vertically the way it's normally displayed?

And like Ameer, I certainly don't understand what that would mean for the bars of a histogram count/pdf/cdf/whatever to start at 0.4.

[Mos_bad]

Actually, as you may see below, my final goal is to plot PDF of different set of data in horizontal direction at the same figure.

[Ameer Hamza]

Do you want to show the pdf as bars or curves? For curves, it can be quite straightforward. For histograms, I couldn't think of a direct way. You will need to add several invisible axes and try to adjust their positions manually. Although possible, only try if it is necessary.

[Image Analyst]

Would the waterfall plot be okay? Check out the waterfall function.

[Mos_bad]

As the figure is 2D, waterfall doesn't work. What I wnt is 2D plot of the horizontal PDF curve for each column vector of dotted data.

Answer 1

[Adam Danz]

Here's a demo that shows how to compute a probability density estimate for each column of data and how to plot the PDF curve next to each column of data. See inline comments for details.

% Produce a matrix of data where each column of y contains
% data for each vertical stack of dots with x-coordinates 
% defined by x.  
x = 0:5:50; 
y = (randn(50, numel(x)) + linspace(.5,5,numel(x))) .* linspace(.5,2,numel(x));

% Plot the data
plot(x, y, 'bo')

% Compute the probability density estimates (pdfx) for each column of y
nSets = size(y,2); 
nPoints = 1000; 
points = linspace(min(y(:)), max(y(:)), nPoints); 
pdfx = cell2mat(arrayfun(@(col){ksdensity(y(:,col),points)'}, 1:nSets)); 

% Compute the interval between the x-values
xInterval = mean(diff(x)); % ie, = 5

% Normalize results so that all pdf values are 
% between 0 and 85% of the x-interval while maintaining 
% the relative heights across all pdf curves.
pdfxNorm = (pdfx - min(pdfx,[],'all')) * (xInterval*.85/range(pdfx(:)));
% NOTE: if you want each pdf curve's height to be independent 
% and *not* maintaining the relative hights across all pdf
% curves, use this line instead:
%   pdfxNorm = (pdfx - min(pdfx,[],1)) .* (xInterval*.85./range(pdfx,1));

% Horizontally offset the pdfxNorm values so each column
% corresponds to an associated x-value.
pdfxShift = pdfxNorm - min(pdfxNorm,[],1) + x;

% Compute y-values for the pdf curves so that the 
% curves do not extend beyond the range of the 
% data within each column. 
pdfYvals = repmat(points(:),1,size(pdfxShift,2)); 
pdfYvals(pdfYvals < min(y, [], 1) | pdfYvals > max(y, [], 1)) = NaN; 

% Add PDFs to plot
axis tight
xlim([min(x), max(x)+xInterval])
hold on
plot(pdfxShift, pdfYvals, 'r-')
grid on

Here's an example of independent-height option for the pdfxNorm variable, using different random data.

[Mos_bad]

Thanks Adam for the help.