Table: add pairPlot function

matStats/@Table/corrcoef.m

@@ -47,7 +47,7 @@
 
 % ------
 % Author: David Legland
-% e-mail: david.legland@inra.fr
+% e-mail: david.legland@inrae.fr
 % Created: 2012-01-10,    using Matlab 7.9.0.529 (R2009b)
 % Copyright 2012 INRA - Cepia Software Platform.
 
@@ -77,7 +77,7 @@
     % correlation coefficient matrix
     mat = corrcoef(data1, data2);
 
-    % foramt into a 2-by-2 Table 
+    % format into a 2-by-2 Table 
     names = {name1, name2};
     res = Table(mat, names, names);
 end

matStats/@Table/correlationCircles.m

@@ -21,7 +21,7 @@
 %     correlationCircles(tab)
 %
 %   See also
-%   corrcoef, plotmatrix
+%     corrcoef, pairPlot, plotmatrix
 %
 %   References
 %   It is inspired by the contribution of Taiyun Wei for R:
@@ -30,7 +30,7 @@
 
 % ------
 % Author: David Legland, based on plotmatrix code
-% e-mail: david.legland@inra.fr
+% e-mail: david.legland@inrae.fr
 % Created: 2012-07-16,    using Matlab 7.9.0.529 (R2009b)
 % Copyright 2012 INRA - Cepia Software Platform.
 

matStats/@Table/pairPlot.m

@@ -0,0 +1,184 @@
+function pairPlot(obj, varargin)
+% Pairwise scatter plots and histograms of table columns.
+%
+%   pairPlot(TAB)
+%   Pair-wise plot of the columns in the specified table.
+%
+%   pairPlot(TAB, GRP)
+%   Pair-wise plot of the columns in the specified table, using a
+%   coloration by group.
+%
+%   pairPlot(..., PNAME, PVALUE)
+%   Specifies additional parameter name-vale pairs. Parameters can be:
+%   * 'Colors':     the color associated to each group, as a NG-by-3 array
+%   * 'HistMode':   the display mode of the histogram on the diagonal. Can
+%                   be one of {'Histogram', 'Bar', 'kde'}.
+%
+%   Examples
+%     % Display pair-wise plot of a simple table
+%     iris = Table.read('fisherIris');
+%     figure;
+%     pairPlot(iris(:,1:4));
+%     % Same display using different color per group
+%     figure;
+%     colors = parula(3);
+%     pairPlot(iris(:,1:4), iris(:,5), 'Colors', colors);
+%
+%   See also
+%     correlationCircles, plot, histogram, violinPlot
+%
+%   References:
+%   Rewritten from the function 'pairplot', by Ryosuke Takeuchi on the
+%   FileExchange
+%   https://fr.mathworks.com/matlabcentral/fileexchange/60866-pairplot-meas-label-group-colors-mode
+%
+
+% ------
+% Author: David Legland
+% e-mail: [email protected]
+% INRAE - BIA Research Unit - BIBS Platform (Nantes)
+% Created: 2020-12-21,    using Matlab 9.8.0.1323502 (R2020a)
+% Copyright 2020 INRAE.
+
+
+%% Parse input arguments
+
+% Check if group argument is provided
+groups = {};
+if nargin > 1 
+    var1 = varargin{1};
+    if isa(var1, 'Table')
+        [groups, groupNames] = parseGroupInfos(var1);
+        groups = groupNames(groups);
+        nGroups = length(groupNames);
+        varargin(1) = [];
+
+    elseif iscell(varargin{1})
+        groups = var1;
+        groupNames = unique(groups);
+        nGroups = length(groupNames);
+        varargin(1) = [];
+    end
+end
+
+% parse group colors
+colors = lines(length(unique(groups)));
+if ~isempty(varargin) &&  isnumeric(varargin{1})
+    colors = varargin{1};
+    varargin(1) = [];
+end
+
+% parse optional arguments
+[colors, varargin] = parseInputOption(varargin, 'Colors', colors);
+[histMode, varargin] = parseInputOption(varargin, 'HistMode', 'histogram'); %#ok<ASGLU>
+
+
+%% Pre processing
+
+% retrieve data and feature names
+data = obj.Data;
+nc = size(data, 2);
+labels = obj.ColNames;
+
+
+%% Scatter plots
+for i = 1:nc
+	for j = 1:nc
+        % select appropriate axis
+		ax = subplot(nc, nc, sub2ind([nc nc], i, j));
+
+        if i == 1
+            ylabel(labels{j});
+        end
+        if j == nc
+            xlabel(labels{i});
+        end
+
+		hold on;
+
+        if i == j
+            continue;
+        end
+
+        if isempty(groups)
+            plot(ax, data(:, i), data(:, j), '.');
+        else
+            for g = 1:nGroups
+                inds = strcmp(groups, groupNames{g});
+                plot(ax, data(inds, i), data(inds, j), ...
+                    '.', 'Color', colors(g,:));
+            end
+        end
+        xlim([min(data(:, i)) max(data(:, i))])
+	end
+end
+
+%% Plot histograms
+for i = 1:nc
+    % select appropriate axis
+    ax = subplot(nc, nc, sub2ind([nc nc], i, i));
+    hold on;
+
+    bins = linspace(min(data(:,i)), max(data(:,i)), 20);
+
+    if strcmpi(histMode, 'histogram')
+        % Display histograms
+        if isempty(groups)
+            histogram(ax, data(:, i), bins, 'Normalization', 'probability');
+        else
+            for g = 1:nGroups
+                inds = strcmp(groups, groupNames{g});
+                histogram(ax, data(inds, i), bins, 'FaceColor', colors(g,:), ...
+                    'Normalization', 'probability');
+            end
+        end
+        xlim([bins(1) bins(end)]);
+
+    elseif strcmpi(histMode, 'bar')
+        % Display bars
+
+        % convert bin centers to edges
+        db = diff(bins) / 2;
+        edges = [bins(1)-db(1), bins(1:end-1)+db, bins(end)+db(end)];
+        edges(2:end) = edges(2:end) + eps(edges(2:end));
+
+        if isempty(groups)
+            [counts, ~] = histcounts(data(:, i), edges);
+            bar(ax, bins, counts, 'BarWidth', 1, 'FaceColor', colors(g,:))
+        else
+            for g = 1:nGroups
+                inds = strcmp(groups, groupNames{g});
+                [counts, ~] = histcounts(data(inds, i), edges);
+                bar(ax, bins, counts, 'BarWidth', 1, 'FaceColor', colors(g,:))
+            end
+        end
+        xlim([edges(1) edges(end)]);
+
+    elseif strcmpi(histMode, 'kde')
+        % Use kernel-density estimate
+        if isempty(groups)
+            [f, xf] = ksdensity(data(:,i));
+            plot(ax, xf, f);
+        else
+            for g = 1:nGroups
+                inds = strcmp(groups, groupNames{g});
+                [f, xf] = ksdensity(data(inds,i));
+                plot(ax, xf, f, 'Color', colors(g,:));
+            end
+        end
+
+    elseif strcmpi(histMode, 'cdf')
+        % Display cumulative density functions
+        if isempty(groups)
+            [f, x] = ecdf(data(inds, i));
+            plot(ax, x, f);
+        else
+            for g = 1:nGroups
+                inds = strcmp(groups, groupNames{g});
+                [f, x] = ecdf(data(inds, i));
+                plot(ax, x, f, 'Color', colors(g,:));
+            end
+        end
+    end
+end
+

tests/test_pairPlot.m

@@ -0,0 +1,40 @@
+function tests = test_pairPlot
+% Test suite for the file pairPlot.
+%
+%   Test suite for the file pairPlot
+%
+%   Example
+%   test_pairPlot
+%
+%   See also
+%     pairPlot
+
+% ------
+% Author: David Legland
+% e-mail: [email protected]
+% Created: 2020-12-22,    using Matlab 9.8.0.1323502 (R2020a)
+% Copyright 2020 INRAE - BIA-BIBS.
+
+tests = functiontests(localfunctions);
+
+function test_Iris(testCase) %#ok<*DEFNU>
+% Test call of function without argument.
+
+iris = Table.read('fisherIris');
+hFig = figure;
+
+pairPlot(iris(:,1:4));
+
+close(hFig);
+
+
+function test_Iris_bySpecies(testCase) %#ok<*DEFNU>
+% Test call of function without argument.
+
+iris = Table.read('fisherIris');
+hFig = figure;
+
+pairPlot(iris(:,1:4), iris(:,5));
+
+close(hFig);
+