How to cluster data in a boxplot?

I am trying to cluster boxplots with no success.

My data is sorted as follows: columns 1 to 8 are subject properties; columns 9 to end are 4 repeated measures of different variables, each column represents a single measure.

I would like to split the file based on the unique combinations of 2 properties (Gait, direction) and cluster the boxplots based on another property (Group). Therefore I would get 4 repeated measures on the x-Axis, its corresponding value on the y-axis and both groups clustered in the plot.

So far I have achieved to plot separately the unique combination of gait, direction and group. I now want to remove group from the combination and use it as a clustering factor.

Can anyone help with this?? thanks in advance!

Data is attached. My code:

numColumns = size(data, 2);  % Total number of columns in the data

% Extract unique gait and direction values
uniqueGaits = unique(data.Gait);
uniqueDirections = unique(data.Direction);
uniqueGroups = unique(data.Group);

% Set the maximum number of open figures
maxOpenFigures = 10;

figureCount = 0;  % Counter for the number of open figures

% Create a dedicated folder for saving figures
folderName = 'Figures';
if ~exist(folderName, 'dir')
    mkdir(folderName);
end

for i = 1:numel(uniqueGaits)
    gait = uniqueGaits{i};
    
    for j = 1:numel(uniqueDirections)
        direction = uniqueDirections{j};
        
        for k = 1:numel(uniqueGroups)
            group = uniqueGroups{k};
            
            % Filter the data based on the unique combination
            filteredData = data(strcmp(data.Gait, gait) & strcmp(data.Direction, direction) & strcmp(data.Group, group), :);
            
            % Iterate over the desired range of column indices
            for col = 9:4:numColumns
                % Extract columns for the current combination
                variableData = table2array(filteredData(:, col:col+3));
                
                % Close figures exceeding the maximum limit
                if figureCount >= maxOpenFigures
                    close all;
                    figureCount = 0;
                end
                
                % Create boxplot for the current combination
                figure;
                boxplot(variableData);
                title(['Boxplot - Gait: ', gait, ', Direction: ', direction, ', Group: ', group]);
                xlabel('Variables');
                ylabel('Value');
                
                % Retrieve column names
                columnNames = filteredData.Properties.VariableNames(col:col+3);
                
                % Set x-axis tick labels as column names
                xticklabels(columnNames);
                
                % Save the figure in the dedicated folder
                saveas(gcf, fullfile(folderName, ['figure_', num2str(figureCount), '.png']));
                
                figureCount = figureCount + 1;
            end
        end
    end
end

% Close any remaining open figures
close all;

Answer 1

[Ive J]

what about this?

data = load("data.mat").data;

% just a personal preference: strings are easier to work with
cellCols = varfun(@(x)isa(x, "cell"), data);
data = convertvars(data, table2array(cellCols), @string);

% group based on combinations of "Direction" and "Gait"
gtab = groupsummary(data, ["Direction", "Gait"], @(x){x}, [7, 9:width(data)]);
gtab.Properties.VariableNames = regexprep(gtab.Properties.VariableNames, "^fun1_", ""); % clean variable names

% find variables for plotting
cols = string(gtab.Properties.VariableNames);
idx = endsWith(cols, "_" + digitsPattern(1) + textBoundary("end")); % _1, _2, etc
viz_cols = cols(idx);
viz_patt = unique(viz_cols.erase("_" + digitsPattern(1) + textBoundary("end"))); % unique pattern of variables for plotting

% loop over each combination pairs
for k1 = 1 % 1:height(gtab) %# DEBUG

    % loop over plotting variables
    for k2 = 1 % 1:numel(viz_patt) %# DEBUG

        close gcf force
        % create a visualization table
        idx = startsWith(cols, viz_patt(k2));
        tmp_col = repmat(gtab.Group(k1), sum(idx), 1);
        tmp_value = gtab{k1, cols(idx)}';
        tmp_group = repelem(1:numel(tmp_value), cellfun(@numel, tmp_value));
        viz_tab = table(categorical(vertcat(tmp_col{:})),...
            vertcat(tmp_value{:}), categorical(tmp_group'), ...
            VariableNames=["cluster", "value", "group"]);

        % colour by 'Group' column and group by repeated measures
        ax = boxchart(viz_tab.group, viz_tab.value, GroupByColor=viz_tab.cluster);
        legend(ax); % change aesthetic of 'ax' if needed 
        xlabel(ax(1).Parent, viz_patt(k2).replace("_", "-"))
        title(ax(1).Parent, sprintf("Direction:%s, Gait:%s", gtab.Direction(k1), gtab.Gait(k1)))
        
        % exportgraphics(gcf, viz_patt(k2) + "." + gtab.Direction(k1) + "."
        % + gtab.Gait(k1) + ".png") %# DEBUG
    end
end

[Marc Elmeua]

This is very helpful, thanks a lot.

I'll try to understand everything in the code, but it works :')