f_colorspaceRepBars.m~

function f_colorspaceRepBars(matches, roundvec, dataname)

% function colorspaceRepBars(matches, binning, dataname)
% meant to show you the ratio of observed matches to expected matches for a
% given bin in rgb space given the assumption of uniform random matches
% matches are nx26x3 matrix of matches
% roundvec bins the rgb space into managable size
% dataname puts a name on the figure and could be used for saving



% useful for plots
letters = {'A' 'B' 'C' 'D' 'E' 'F' 'G' 'H' 'I' 'J' 'K' 'L' 'M' 'N' 'O' 'P' 'Q' 'R' 'S' 'T' 'U' 'V' 'W' 'X' 'Y' 'Z'};


% counts of letters in each rounded 3d bin across all letters
allletterhists = zeros((length(roundvec))^3,1);

% rounded rgb values across all letters
allroundedclrs = [];
clrctr = 1;

% for each letter
for i=1:26
    clrs = squeeze(matches(:,i,:));
    %  all rgb values are between 0 and 1
    % we could round to the nearest .1
    % use downloaded function roundtowardvec
%     roundvec = 0:.25:1; %(1331 locations)
    roundedclrs = roundtowardvec(clrs,roundvec);
    
    allroundedclrs = [allroundedclrs; roundedclrs];
    
    % want to be able to scale our plots (either alpha or markersize) using the
    % number of datapoints in each bin.  3d hist code in matlab looks like a
    % pain so we could do it in a loop?
    
    counter=1;
    bplotmatrix = zeros(length(roundvec)^3,4);
    
    for a=1:length(roundvec)
        for b=1:length(roundvec)
            for c=1:length(roundvec)
                bplotmatrix(counter,:) = [roundvec(a) roundvec(b) roundvec(c) ...
                    sum(ismember(roundedclrs, [roundvec(a) roundvec(b) roundvec(c)],...
                    'rows'))];
                counter=counter+1;
            end
        end
    end
    % less than a minute
    
%     collect matches for this letter for all letters plot
     allletterhists = allletterhists + bplotmatrix(:,4);
    
    
%     uniform matches at each point
      expectedmatches = sum(bplotmatrix(:,4))/(length(roundvec))^3;
    
%     for the bar plot we need to sort the bplotmatrix
    [sorted sortindex] = sort(bplotmatrix(:,4));
    
    
%     so for each letter if we want it later

% figure('name',['distribution across rgb space for ' dataname ' ' letters{i}],'Color',[1 1 1]);
%     hBar=bar(1:(length(roundvec))^3,bplotmatrix(sortindex,4)/expectedmatches,'hist');
%     box off
% %     set(gca,'XTickLabel',names,'XLim',[-0.5,10.5],'YLim',[0,.5]);
%     ylabel('% of responses');
%     set(hBar,'FaceVertexCData',bplotmatrix(sortindex,1:3));
%    
%     
  
    
end



% across all letters

   
%     uniform matches at each point
      expectedmatches = sum(allletterhists)/(length(roundvec))^3;
    
%     for the bar plot we need to sort the bplotmatrix
    [sorted sortindex] = sort(allletterhists);
    
    
%     so for each letter

figure('name',['distribution across rgb space for ' dataname ' all letters'],'Color',[1 1 1]);
%   trick for having bars and lines in the same plot is to have two axes
%   but this yyaxis command is only in matlab 2016a
%     yyaxis 'left

    hBar=bar(1:(length(roundvec))^3,allletterhists(sortindex)/expectedmatches,'hist');
    box off
%     set(gca,'XTickLabel',names,'XLim',[-0.5,10.5],'YLim',[0,.5]);
    ylabel('ratio of matches to uniform');
    set(hBar,'FaceVertexCData',bplotmatrix(sortindex,1:3));
    hold on;
    set(gca,'XLim',[0.5 length(roundvec)^3+.5]);
%     save current axis
    a1 = gca;
%     make a new one in exactly the same place
    a2 = axes('Position',get(a1,'Position'));
 %     this should add a dashed line at 1 meaning chosen proportionally to
%     area of space (or area, diam, or whatever is 1)
    plot(0:.5:length(roundvec)^3,1,'k-');
    hold on;

%     make second axis invixible
  set(a2,'YAxisLocation','right','Color','none','XTickLabel',[])

% Align the x-axis of both axes and display the grid lines on top of the bars.

    set(a2,'XLim',get(a1,'XLim'),'Layer','top')

    
%     le