Homec4science
Diffusion Research Scripts (Thomas Bolton) (master) Browse

Scripts_Sharing/SCHMM/Plotting/CreateGraph_JAJA.m
bdfe7aadccccmaster

Last Change
Raw File

CreateGraph_JAJA.m
View Options

%% This function creates a reasonably fancy graph representation, from an
% adjacency matrix W
%
% Inputs:
% - W
% - is_dir is a logical scalar denoting whether the provided information is
% directional or not (typically, enter '1' for SCHMM data)
% - T is the threshold at which edges are not displayed
% - Size is the 'baseline' size of nodes in the display (on top of which we
% add with the adjacency information)
% - CMin and CMax are the color thresholds for the display, to tune given
% the desired output
% - G_Std contains the information about standard deviation for a given
% edge, to represent its robustness by the edge color. Enter an empty
% matrix if you do not wish to represent that information
% - is_inout is a logical scalar denoting whether nodes should have a color
% matching their directional or symmeterized influence
% - weight_edges is a tuning parameter for the width of the edges
% - is_MNM is a logical scalar denoting whether you would like node color
% to depict statistical significance of incoming vs outgoing edges for the
% nodes (across bootstrapping folds)
% - MNM is the related data to enter if you wish to do so (a 3D adjacency
% of size n_network x n_network x n_bootstrap_folds
% - COLMAP is the colormap to use (string entry, e.g. 'jet')
%
% Outputs:
% - G is the created graph
% - Std_min and Std_max reflect STD boundaries for the considered graph
function [G,Std_min,Std_max] = CreateGraph_JAJA(W,is_dir,T,Size,CMin,CMax,G_Std,is_inout,weight_edges,is_MNM,MNM,COLMAP)
% Trims the adjacency matrix for edges weaker than T
W_full = W;
W(W < T) = 0;
%
if ~isempty(G_Std)
G_Std(W < T) = 0;
end
% Creates the graph and associated propertes
if is_dir
G = digraph(W);
if is_inout
G.Nodes.NodeColors = sum(W_full,2);
else
G.Nodes.NodeColors = sum(W_full,2) + sum(W_full,1)';
end
else
G = graph(W);
G.Nodes.NodeColors = sum(W_full,1)';
end
% Sets the width of the edges proportional to the actual values
G.Edges.LWidths = G.Edges.Weight;
G.Edges.LWidths = weight_edges*G.Edges.LWidths;
h = plot(G,'Layout','circle');
set(gcf,'color','w');
set(gca,'Box','off');
axis off;
% If it is desired to plot node colors to reflect statistical
% significance of ingoing vs outgoing nodal influences, that is
% computed and used as NodeCData
if is_MNM
tmp_data = squeeze(sum(MNM,2));
tmp_data2 = squeeze(sum(MNM,1));
for i = 1:size(W,1)
[~,~,~,stats] = ttest2(tmp_data(i,:)',tmp_data2(i,:)');
tmp_NODE(i) = stats.tstat;
end
h.NodeCData = tmp_NODE;
else
h.NodeCData = G.Nodes.NodeColors;
end
h.LineWidth = G.Edges.LWidths;
if is_inout
h.MarkerSize = Size + Size*(sum(W,1)');
else
h.MarkerSize = Size + Size*G.Nodes.NodeColors;
end
h.NodeLabel = {};
% Puts the graph information in the plot display, either with gray
% edges (no STD info available) or with colored edges otherwise
if isempty(G_Std)
h.EdgeColor = [0.5 0.5 0.5];
else
tmp_val = G_Std(W > 0);
Std_min = min(tmp_val);
Std_max = max(tmp_val);
Delta_Std = Std_max - Std_min;
Delta_CM = 1000 - 1;
idx_CM = (Delta_CM/Delta_Std)*tmp_val + (1 - (Delta_CM/Delta_Std)*Std_min);
min(idx_CM)
max(idx_CM)
CM = flipud(jet(1000));
h.EdgeColor = CM(round(idx_CM),:);
end
if is_dir
h.ArrowSize = 12;
end
h.EdgeAlpha = 0.3;
colormap(COLMAP);
caxis([CMin,CMax]);
end
c4science · Help