boxplot from a structured array

hi

i have a structured array it's like this

name_of_image(1).stats
name_of_image(2).stats
etc

in .stats is a whole load of variables, like stats.area

what is the simplest way of producing a box plot of e.g stats.area, while ensuring that the box plot is labeled correctly with the name_of_image

thanks

ps i tried a for loop to extract the information, but thought there must be a simpler way

x = []; 
% nameArray = [] ; 
for i = 1 : length(collated_data)

% nameArray(i) = collated_data(i).name; 

      for k = 1 : length(collated_data(i).stats)
          x(i, k) = collated_data(i).stats(k).area; 
          if x(i, k) < 17 
             x(i, k) = NaN; 
          end
      end

end

[the cyclist]

Can you upload a MAT file with this variable in it?

[cgenes]

here it is

[the cyclist]

FYI, in your code the variable is named "area", but in the file it is "Area". So, you might need to fix that up in my code.

Answer 1

[the cyclist]

numberArrays = numel(collated_data);

areaAndGroup = [];

for na = 1:numberArrays
    numberAreas = numel(collated_data(na).stats);
     % This is the tricky part. I'm extracting the area data from the inner structure, munging them into a vector, and then including a second
     % column that enumerates which structure they came from (to use as a grouping variable in the boxplot command).
    areaAndGroup = [areaAndGroup; [collated_data(na).stats.Area]',repmat(na,[numberAreas,1])];
end

areaAndGroup(areaAndGroup(:,1)<17,:) = NaN;

figure
boxplot(areaAndGroup(:,1),areaAndGroup(:,2))

[cgenes]

yes thanks very much for that very nice. But i actually want the names of the images, not the numbers on the x-axis.

I think i'm storing my data wrong - - I should not use a structured array , but instead have a character array for image names, and arrays for variables such as area -- much like the box plot example load carsmall.

if I had image X, and wanted to stored variable (from regionprops) Y and Z, how would I go about storing the data like in the example 'load carsmall'?

PS if this is worthy of another topic I will accept your first answer!! :-)

[the cyclist]

Well, you've got some awfully long names there. The main problem will be parsing those into something that will fit. Here is a not-very-robust attempt:

numberArrays = numel(collated_data);

areaAndGroup = [];

for na = 1:numberArrays
    numberAreas = numel(collated_data(na).stats);
    shortName{na} = collated_data(na).name(41:end-14);
     % This is the tricky part. I'm extracting individual area, munging them into a vector, and then including a second
     % column enumerated which structure they came from (to use as a grouping variable).
    areaAndGroup = [areaAndGroup; [collated_data(na).stats.Area]',repmat(na,[numberAreas,1])];
end

areaAndGroup(areaAndGroup(:,1)<17,:) = NaN;

figure
boxplot(areaAndGroup(:,1),areaAndGroup(:,2))
set(gca,'XTickLabel',shortName,'XTickLabelRotation',30)

You could do something smarter using regular expressions to identify the relevant part of the longer name.

[cgenes]

thanks very much