FG_RTdata.m

%% (1) basics - loading data, extracting variables of interest, preallocating results array

%subjectN = input('Enter subject number: ', 's');
%subjectN = str2double(subjectN);

% fileS = 'C:\Users\Luca\Documents\mta-ttk anyagok\FG-RT\sampleData\sub2Log.mat';
% fileP = '/home/adamb/FG-RT/sampleData/sub2Log.mat';

%subID = sprintf('sub%dLog.mat', subjectN);
%load('subID');
load('sub2Log.mat');

RT_all = cell2mat(logVar(2:end,12));
blockIndices = cell2mat(logVar(2:end, 2));
figPresent = cell2mat(logVar(2:end, 7));  % binary vector
stim_difficulty = cell2mat(logVar(2:end,5));
accuracy = logical(cell2mat(logVar(2:end,10))); % need the xtra 'logical' step (?)

% somehow extract the easy-difficult distinction in a general way:
diffValues = [min(unique(stim_difficulty)), max(unique(stim_difficulty))];
isDifficult = stim_difficulty==diffValues(2);  % binary vector

% what will our results look like
subRT = nan(20, 10, 2, 2);


%% (2) sort RTs into categories, store it in output variable

for blockIdx = 1:10
    
    for fig = 0:1
        
        for diff = 0:1
            
            subRT(:, blockIdx, fig+1, diff+1) = RT_all(blockIndices==blockIdx & figPresent==fig & isDifficult==diff);
            
        end
    end
    
end

%%%%%%%%%%%%% adam's comments here %%%%%%%%%%%%%
% alternative method for merging blocks:
subRTreshaped = reshape(subRT, [200, 2, 2]);  % elements of input are taken columnwise, working out nicely in our case

% get mean and SD for RT per condition
mean_stmType = mean(subRTreshaped, 'omitnan');  % if there is even one NaN value, the result would be NaN, so we use the "omitnan" flag
sd_stmType = std(subRTreshaped, 'omitnan');
%%%%%%%%%%%%% adam's comments ended %%%%%%%%%%%%


% For me the line below gives a 2x2x2 matrix - was that the intention?

%%%%%%%% if subRT_types is 4D, Mean_SD_stmType gives also a 4D, only
%%%%%%%% created 4D for readability issues --> 2x2x2 matrix separates only
%%%%%%%% by difficulty (:,:,1) or (:,:,2)

% first row mean, second row SD
Mean_SD_stmType = vertcat(mean_stmType, sd_stmType);


%% (3) optional - report some basic stuff:
% mean & SD of RT for the 2 X 2 structure for the overall task (all blocks combined)
% mean & SD of RT for the 2 X 2 structure per block (+ per stimulus type)

mean_RT = mean(subRT(:), 'omitnan'); % omit NAN values
sd_RT = std(subRT(:), 'omitnan');


%%%%%%%%%%%%% adam's comments here %%%%%%%%%%%%%

% You already have all RT data arranged nicely according to blocks in var
% "subRT". You can simply take the means / SDs along the appropriate
% dimension to obtain block-level mean and SD:

RTblockMean = mean(subRT, 1, 'omitnan');  % mean along first dimension, where size is 20
RTblockSD = std(subRT, 0, 1, 'omitnan');  % std along first dimension - Note the extra input argument!

% Since our input data might contain NaN values (where there was a missing
% answer), we need to use the "omitnan" flag, otherwise we can easily end
% up with NaN as the result.

RTblockMeanSD = vertcat(RTblockMean, RTblockSD);

%%%%%%%%%%%%% adam's comments ended %%%%%%%%%%%%

%% Accuracy data - do we need hit/miss, false alarm/correct rejection rates?

Num_acc = numel(accuracy(accuracy==1));
proportion_acc = Num_acc/800*100;

subAcc = nan(20, 10, 2, 2);

for blockIdx = 1:10
    
    for fig = 0:1
        
        for diff = 0:1
            
            subAcc(:, blockIdx, fig+1, diff+1) = accuracy(blockIndices==blockIdx & figPresent==fig & isDifficult==diff);
            
        end
    end   
end

MeanAccuracy_block = mean(subAcc, 'omitnan');
subAccReshaped = reshape(subAcc, [200,2,2]);
MeanAccuracy = mean(subAccReshaped, 'omitnan');

%% mean & SD for accurate/inaccurate

% mean_acc = mean(RT_all(accuracy==1));
% sd_acc = std(RT_all(accuracy==1));
% mean_inacc = mean(RT_all(accuracy==0));
% sd_inacc = std(RT_all(accuracy==0));

%% per blocks and stim type

% duplicate the subRT matrix to avoid loss of data 
subRTdup1 = repmat(subRT,1);
subRTdup2 = repmat(subRT,1);

% assign nan for the inaccurate trials to leave them out
subRTdup1(accuracy==0) = nan;
blockMean_acc = mean(subRTdup1, 'omitnan');
blockSD_acc = std(subRTdup1, 'omitnan'); 

% now for the accurate trials
subRTdup2(accuracy==1) = nan;
blockMean_inacc = mean(subRTdup2, 'omitnan');
blockSD_inacc = std(subRTdup2, 'omitnan'); 


%% all blocks, per stim type

accReshaped1 = reshape(subRTdup1, [200,1,2,2]);
accReshaped2 = reshape(subRTdup2, [200,1,2,2]);

meanAcc = mean(accReshaped1, 'omitnan');
sdAcc = std(accReshaped1, 'omitnan');
meanInacc = mean(accReshaped2, 'omitnan');
sdInacc = std(accReshaped2, 'omitnan');


%% Mean and SD per blocks (1-10) and overall by stimlus type (11th column)

%%%%%%%%%%%%% adam's comments here %%%%%%%%%%%%%
% The line below doesn't work for me
%%%%%%%%%%%%% adam's comments ended %%%%%%%%%%%%

%%%%%% it only works if Mean_SD_stmType is also a 4D matrix, it matches in
%%%%%% size (not at all neccessary though)

% Behav_all = horzcat(Mean_SD_blocks, Mean_SD_stmType);


%% Plots
% RT and Accuracy per blocks and stimulus type

figure(1)
hold on
for fig = 0:1
    
    for diff = 0:1
        
        plot(RTblockMean(:,:,fig+1,diff+1), 'o-');  % added circles for each "real" data point
        
    end
end

legend('fig absent / easy', 'figure absent / diff', 'figure present / easy', 'figure present / diff');
xlabel('Block no.');
ylabel('Mean RT (ms)');

%%%%%%%%%%%%% adam's comments here %%%%%%%%%%%%%
% For the bar graph:
% The bar graph could take as input a matrix and then it would group
% conditions automatically, according to the matrix structure. And we turn
% the var "mean_stmType" into a 2x2 matrix and just feed it to the bar()
% function

figure(2);
% squeeze and transpose the data so that we group primarily by
% Easy/Difficult categories
tmpData = squeeze(mean_stmType); tmpData = tmpData';
b = bar(tmpData);
% add labels to bars
set(gca,'XTickLabel', {'Easy', 'Difficult'});
% add legend to specify Figure Absent / Present categories
legend('Figure Absent', 'Figure Present');
xlabel('Stimulus types');
ylabel('Mean RT (ms)');

%%%%%%%%%%%%% adam's comments ended %%%%%%%%%%%%

figure(3);
hold on
for fig = 0:1
    
    for diff = 0:1
        
        plot(MeanAccuracy_block(:,:,fig+1,diff+1), 'o-');
        
    end
end
        
legend('fig absent / easy', 'figure absent / diff', 'figure present / easy', 'figure present / diff');
xlabel('Block no.');
ylabel('Correct responses (%)');       
        
figure(4);
tmpData2 = squeeze(MeanAccuracy); tmpData2 = tmpData2';
b2 = bar(tmpData2);
set(gca,'XTickLabel', {'Easy', 'Difficult'});
legend('Figure Absent', 'Figure Present');
xlabel('Stimulus types');
ylabel('Correct responses (%)');