makePlotsForEmpiricalPaper.m

function makePlotsForEmpiricalPaper(varargin)
% By default plots produced will be saved in a subfolder of the folder in
% which this function is located.

% INPUT
% varargin{1}: String. Plot using the 'mean' or 'median' accross 
%   participants. Deafault is 'mean'. The 'median' option is currently
%   untested and has not been carefully checked. Always uses SEM so 
%   'median' doesn't really make sense. When 'median' is selected only a 
%   subset of plots is done.
% varargin{2}: String. Provide a directory to save in a different location
%   to usual. No trailing slash.

baseCodeDir = mfilename('fullpath');
[baseCodeDir, ~, ~] = fileparts(baseCodeDir);

if (~isempty(varargin)) && (~isempty(varargin{1}))
    plotStat = varargin{1};
else
    plotStat = 'mean';
end

if (length(varargin)>1) && (~isempty(varargin{2}))
    saveDir = varargin{2};
else
    saveDir = fullfile(baseCodeDir, 'PlotsAndResults');
    mkdir(saveDir)
end

baseDataDir = load(fullfile(baseCodeDir, 'confDataAndFitDir.mat'));
baseDataDir = baseDataDir.dataDir;

RealData = fullfile(baseDataDir, 'RealData', '_standardFormatData.mat');
CrossValidationFits = fullfile(baseDataDir, 'CrossValidationFits');
FullDataSetFits = fullfile(baseDataDir, 'FullDataSetFits');
RegularisedFits = fullfile(baseDataDir, 'RegularisedFits');
RegularisedCvFits = fullfile(baseDataDir, 'RegularisedCvFits');
modelRecovFolder = fullfile(baseDataDir, 'ModelRecov');

modelNames = {'0', 'V', 'D', 'VD', 'VC', 'VDC', 'M', 'VM', 'DM', 'VDM'};
origNames = {   'NDscFlatNoneSameMvar',...
                'NDscFlatDvarSameMvar',...
                'NDscSlpeNoneSameMvar',...
                'NDscSlpeDvarSameMvar',...
                'TrDsFlatDvarSameMvar',...
                'TrDsSlpeDvarSameMvar',...
                'FaDsFlatNoneSameMvar',...
                'FaDsFlatDvarSameMvar',...
                'FaDsSlpeNoneSameMvar',...
                'FaDsSlpeDvarSameMvar'};


%% Illustrations
dispMsg('Illustrations')

bayesianConfidenceIllustration('rawEv');
mT_exportNicePdf(15.9, 15.9, saveDir, 'confIllustration_rawEv')

bayesianConfidenceIllustration('pCorrect');
mT_exportNicePdf(15.9, 15.9, saveDir, 'confIllustration_pCorrect')

bayesianConfidenceIllustration('logPostRatioForChoice');
mT_exportNicePdf(15.9, 15.9, saveDir, 'confIllustration_logPostRatio')

bayesianConfidenceIllustration('logPostRatioForChoice2');
mT_exportNicePdf(15.9, 15.9, saveDir, 'confIllustration_logPostRatio2')

bayesianConfidenceIllustration('logPostRatioForChoice', 'full');
mT_exportNicePdf(15.9, 15.9, saveDir, ...
    'confIllustration_logPostRatio_fullRange')

bayesianConfidenceIllustration('logPostRatioForChoice2', 'full');
mT_exportNicePdf(15.9, 15.9, saveDir, ...
    'confIllustration_logPostRatio2_fullRange')

figure
x = -100:0.01:100;
y = normpdf(x, 0, 20);
plot(x, y)
mT_exportNicePdf(15.9, 15.9, saveDir, 'confIllustration_normalDist')


%% Basic data plots
dispMsg('Real data plots')

if strcmp(plotStat, 'mean')
    LoadedFiles = load(RealData);
    DSet = LoadedFiles.DSet;

    plotDencity(DSet)
    mT_exportNicePdf(15.9/2, 15.9, saveDir, 'expB_dencities')
    close all
end

LoadedFiles = load(RealData);
DSet = LoadedFiles.DSet;
DSet = prepDataWrapper(DSet, 4, true);

figHandle = figure;
for plotType = {'scatter', 'line'}
    figHandle = plotConfAgainstTimeAndEv(DSet, plotType{1}, figHandle, ...
        'binned', false, false, 'B', 'noEv', [], plotStat, [], true);
end
mT_exportNicePdf(15.9/2, 15.9/2, saveDir, 'confTimeMidpointInteraction')
close all

if strcmp(plotStat, 'mean')
    figHandle = figure;
    plotConfFrequencies(DSet, 'scatter', figHandle);
end

figHandle = figure;
figHandle = plotAccAndConfWithRt(DSet, 'scatter', 'B', false, ...
    figHandle, plotStat);
plotAccAndConfWithRt(DSet, 'line', 'B', false, ...
    figHandle, plotStat);
mT_exportNicePdf(15.9, 15.9/2, saveDir, 'expB_combinedTime')
close all

% Plots but looking at raw confidence instead of binned conf
f = figure;
plotConfAgainstTimeAndEv(DSet, 'scatter', f, 'raw', false, false, 'B')
mT_exportNicePdf(15.9, 15.9/2, saveDir, 'expB_rawConfEffects')
close all


%% Qualitative analysis
dispMsg('Quantiative analysis')
if strcmp(plotStat, 'mean')
    [medianNonDecis, evEffectFig] = runStatistics(DSet);
    figure(evEffectFig)
    mT_exportNicePdf(15.9/2, 15.9*(7/12), saveDir, 'expB_evEffectCoefs')
    close all
end


%% Evidence residuals
dispMsg('Evidence residuals')
EvDSet = convertToEvSet(DSet, true);

% Shade the region corresponding to non-decision time (Stim frame duration 
% was 50ms)
shadedRegion = [-medianNonDecis*0.05, 0];

plotFig = plotEvResidualsEffectWithTime(EvDSet, 'scatter', 'B', ...
    true, false, [], [], plotStat, shadedRegion);
plotFig = plotEvResidualsEffectWithTime(EvDSet, 'line', 'B', ...
    true, false, plotFig, [], plotStat);
figure(plotFig)
mT_exportNicePdf(15.9, 15.9, saveDir, 'expB_combinedEv')


%% Modelling results 
dispMsg('Modelling main overview')

if strcmp(plotStat, 'mean')
    [~, crossValPlot, CvDSets] = runCrossValEval(CrossValidationFits, false, ...
        true, modelNames, false, true);
    assert(isequal(origNames, mT_findAppliedModels(CvDSets{1})))

    figure(crossValPlot)
    mT_exportNicePdf(15.9/2, 15.9, saveDir, 'crossValComparison')
end

[DSet, FigureHandles] = mT_analyseClusterResults(FullDataSetFits, ...
    1, true, false, true, modelNames); 
DSet = DSet{1};
assert(isequal(origNames, mT_findAppliedModels(DSet)))

[~, bicData] = mT_collectBicAndAicInfo(DSet);
runBicBasedMixEffectsModelCompare(bicData, modelNames)

if strcmp(plotStat, 'mean')
    figure(FigureHandles.AicBic)
    mT_exportNicePdf(15.9, 15.9, saveDir, 'aicBic')

    % The threshold parameters are in a random order. Sort to be in
    % order before making the parameter table for latex
    TmpDSet = DSet;
    for iP = 1 : length(TmpDSet.P)
        for iM = 1 : length(TmpDSet.P(iP).Models)
            oldThresh = TmpDSet.P(iP).Models(iM).BestFit.Params.Thresholds;
            assert(size(oldThresh, 2) == 1)
            TmpDSet.P(iP).Models(iM).BestFit.Params.Thresholds = ...
                sort(oldThresh);
        end
    end

    ParamLabels.Sigma_phi = "$\sigma_\varphi$";
    ParamLabels.Sigma_acc = "$\sigma_{acc}$";
    ParamLabels.Thresholds = "$d_i$";
    ParamLabels.BoundIntercept = "$a$";
    ParamLabels.PipelineI = "$I$ (s)";
    ParamLabels.LapseRate = "$\lambda$";
    ParamLabels.MetacogNoise = "$\sigma_m$";
    ParamLabels.NoiseRatio = "$\Gamma$";
    mT_produceParamStats(TmpDSet, saveDir, [2, 5, 7], ParamLabels, ...
                            [], modelNames)
    close all

    % Fit end points
    [~, restartsFigure] = mT_plotFitEndPoints(DSet, false, 1, [], modelNames);
    figure(restartsFigure)
    mT_exportNicePdf(15.9/2, 15.9*(2/3), saveDir, 'restartsRequired')
end

%% Main modelling fits
dispMsg('Modelling main fits')

plotAndSaveConfidencePredictions(DSet, 7, plotStat, saveDir, '')

makeCombinedPoorModelsPlot(DSet, [2, 5], modelNames, plotStat, saveDir)

dispMsg('Modelling main full sim')
plotFullSimFits(saveDir, DSet, 7, plotStat, '', 640*10, [], 'skipRtPlots')


%% Regularised fits
dispMsg('Modelling regularised')

[RegDSet, ~] = mT_analyseClusterResults(RegularisedFits, 1, ...
    true, false, true);
RegDSet = RegDSet{1};
close all

applMmodels = mT_findAppliedModels(RegDSet);
iM =  1;
assert(strcmp(applMmodels{iM}, 'FaDsFlatNoneSameMvarReg1'))
plotAndSaveConfidencePredictions(RegDSet, iM, plotStat, ...
    saveDir, ['model_' applMmodels{iM} '_'])

plotFullSimFits(saveDir, RegDSet, iM, plotStat, ...
    ['model_' applMmodels{iM} '_'], 640*5)


%% Regularised CV fits

if strcmp(plotStat, 'mean')
    runCrossValEval(RegularisedCvFits, false, true, [], true);
end


%% Model recovery analyses

if strcmp(plotStat, 'mean')
    dispMsg('Model recovery')
    produceModelRecoveryPlots(modelRecovFolder, modelNames, 7, saveDir, '')
end

end


function dispMsg(msg)
disp('')
disp(['**** ' msg])
end