f_nansdistribution.m

function   nansinfo = f_nansdistribution(rgb_data, group)

%  nansinfo = f_nansdistribution(rgb_data, group)
% rgb_data is subset of u_rgb
% group is a string used for figure labels designating subset of
% synesthetes used
% output
% take a set of rgb matching data and count the number of letters not
% matched in various ways
% total number of possible matches
% number not matched
% number not matched by letter
% number not matched by subject
% 





% this is handy for labeling figures
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'};


% index to subjects with NaNs
subswnans = [];
% how many NaNs for each subject
numnans = [];
% for each subject
for i=1:length(rgb_data)
%     check for a nan
    y=sum(isnan(rgb_data(i,2,:)));
    if y>0
%     fprintf('subject %g has %g nans \n',i,y);
    subswnans = [subswnans i];
    numnans = [numnans y];
    else
        numnans = [numnans 0];
    end
end
    
% % so less than 5% of the matches are NaNs
% 6588*26=171288
%       
% sum(numnans)
% ans =
%         7352
% 7352/171288
% ans =
%      0.042922
% which is not too bad, but we need to account for them.

%  these are distributed throughout 35% of the population (which is more
%  than I would have expected)
% size(subswnans,2)/6588 =0.352

% if we do nothing, the nans get assigned the color black (0,0,0);

% the matches also get the label 0 which corresponds also to the color
% black

% however if we don't allow the interpolation at the label assignment
% stage, then things should fix themselves in the histograms.  this fix is
% in the function fpRGB2ColorsJW.m.   All the functions appear to work now.
% let's make some plots so we can see where the problems are.  


% make a directory to save the figures if one doesn't exist

nansavdir = 'nansdistribution';

if ~exist(nansavdir,'dir')
    mkdir(nansavdir);
end


% number of histogram of subjects with n NaNs

figure('Name','histogram of subjects with n letters with no color','Color',[1 1 1],'Position',get(0,'ScreenSize'));
hist(numnans,[0:26]);
xlabel('number of letters with no match');
% so many are 0 that we need a log scale on y
hold on;
ylabel('number of subjects');
box off;
set(gca,'XLim',[-.5 26]);
saveas(gcf,[nansavdir '/' group '.histofnomatchesAllsubs.png'],'png');
plot2svg([nansavdir '/' group '.histofnomatchesAllsubs.svg'],gcf);

close(gcf);



% do this as percentages
h=hist(numnans,[0:26]);

pctnans = h/sum(h);

% 
%             0        0.648
%             1      0.12614
%             2      0.05935
%             3      0.03901
%             4     0.034153
%             5     0.025956
%             6     0.022769
%             7     0.018367
%             8     0.016697
%             9    0.0069824
%            10    0.0022769
%            11   0.00030358
%            12            0
%            13            0
%            14            0
%            15            0
%            16            0
%            17            0
%            18            0
%            19            0
%            20            0
%            21            0
%            22            0
%            23            0
%            24            0
%            25            0
%            26            0



figure('Name','percent of subjects with n letters with no color','Color',[1 1 1],'Position',get(0,'ScreenSize'));
bar(pctnans);

xlabel('number of letters with no match');
ylabel('percent of subjects');
box off;
set(gca,'XLim',[0 26]);

saveas(gcf,[nansavdir '/' group 'normhistofnomatchesAllsubs.png'],'png');
plot2svg([nansavdir '/' group 'normhistofnomatchesAllsubs.svg'],gcf);
% close(gcf);





% how is this behavior distributed across letters?
figure('Name','distribution of nans across letters','Color',[1 1 1],'Position',get(0,'ScreenSize'));

subplot(1,2,1);
% figure where nans are black and color matches are white
% get one rgb column of data for subs x letters
nansmatrix = squeeze(rgb_data(:,2,:));

nansmatrix(find(~isnan(nansmatrix)))=0;
nansmatrix(find(isnan(nansmatrix)))=1;
imagesc(nansmatrix);  
colormap(bone);
box off;
xlabel('letters');
ylabel('subjects');
set(gca,'XTick',[1:26],'XTickLabel',letters);
% set(gca,'XTickLabel',letters);


% let's look at the distribution across letters
subplot(1,2,2);

bar(sum(nansmatrix)/length(rgb_data));
box off;
xlabel('letters');
ylabel('% of times not matched');
set(gca,'XTick',[1:26],'XTickLabel',letters);
% set(gca,'XTickLabel',letters);
saveas(gcf,[nansavdir '/' group 'distofnonmatchesAcrossLetters.png'],'png');
plot2svg([nansavdir '/' group 'distofnonmatchesAcrossLetters.svg'],gcf);
% close(gcf);

% 
% need to add lewand frequency here if this is going to work. is still in
% main script
% 
% % hmmmm is number of nans predicted by the letter frequency?
% 
% lewand frequency
% 5 20  1 15  9 14 19  8 18  4 12 3 21  13 23 6 7 25  16  2 22   11 10 24 17 26
% e t   a o   i n  s   h r   d l  c u   m  w  f g y p   b v    k  j  x  q  z
fqorder = [5 20  1 15  9 14 19  8 18  4 12 3 21  13 23 6 7 25  16  2 22   11 10 24 17 26];

% how is this behavior distributed across letters?
figure('Name','distribution of nans across letters fqsort','Color',[1 1 1],'Position',get(0,'ScreenSize'));

subplot(1,2,1);
% figure where nans are black and color matches are white
% get one rgb column of data for subs x letters
nansmatrix = squeeze(rgb_data(:,2,:));

nansmatrix(find(~isnan(nansmatrix)))=0;
nansmatrix(find(isnan(nansmatrix)))=1;
imagesc(nansmatrix(:,fqorder));  
colormap(bone);
box off;
xlabel('letters');
ylabel('subjects');
set(gca,'XTick',[1:26],'XTickLabel',letters(fqorder));
% set(gca,'XTickLabel',letters);


% let's look at the distribution across letters
subplot(1,2,2);

bar(sum(nansmatrix(:,fqorder))/length(rgb_data));
box off;
xlabel('letters');
ylabel('% of times not matched');
set(gca,'XTick',[1:26],'XTickLabel',letters(fqorder));
% set(gca,'XTickLabel',letters);
saveas(gcf,[nansavdir '/' group 'distofnonmatchesAcrossLettersFQ.png'],'png');
plot2svg([nansavdir '/' group 'distofnonmatchesAcrossLettersFQ.svg'],gcf);
% close(gcf);

% fq of first letter in a word

% 20 1 19 8 23 9 15 2 13 6 3 12 4 16 14 5 7 18 25 21 22 10 11 24 26 24
% t  a s  h w  i o  b m  f c l  d p  n  e g r  y  u  v  j  k  q  z  x

fqfirst = [20 1 19 8 23 9 15 2 13 6 3 12 4 16 14 5 7 18 25 21 22 10 11 24 26 24];

% how is this behavior distributed across letters?
figure('Name','distribution of nans across letters fqsort','Color',[1 1 1],'Position',get(0,'ScreenSize'));

subplot(1,2,1);
% figure where nans are black and color matches are white
% get one rgb column of data for subs x letters
nansmatrix = squeeze(rgb_data(:,2,:));

nansmatrix(find(~isnan(nansmatrix)))=0;
nansmatrix(find(isnan(nansmatrix)))=1;
imagesc(nansmatrix(:,fqfirst));  
colormap(bone);
box off;
xlabel('letters');
ylabel('subjects');
set(gca,'XTick',[1:26],'XTickLabel',letters(fqfirst));
% set(gca,'XTickLabel',letters);


% let's look at the distribution across letters
subplot(1,2,2);

bar(sum(nansmatrix(:,fqfirst))/length(rgb_data));
box off;
xlabel('letters');
ylabel('% of times not matched');
set(gca,'XTick',[1:26],'XTickLabel',letters(fqfirst));
% set(gca,'XTickLabel',letters);
saveas(gcf,[nansavdir '/' group 'distofnonmatchesAcrossLettersFQfirst.png'],'png');
plot2svg([nansavdir '/' group 'distofnonmatchesAcrossLettersFQfirst.svg'],gcf);
% close(gcf);



% make some tables and return some info

nansinfo.totalmatches = length(rgb_data)*26;
nansinfo.totalnans = sum(numnans);
nansinfo.nansbyletter = sum(nansmatrix);
nansinfo.nansbysubject = hist(numnans,0:26);







end

% 
% [nancor nanp] = corr(log(lewandfq),pctnansbyletter','rows','pairwise','type','spearman');
% figure;
% 
% plot(log(lewandfq),pctnansbyletter','ro');
% hold on;
% text(log(lewandfq),pctnansbyletter',letters);