Multiple shaded spans on a time series plot dependent on a condition for separate time series data

Hi, I need to add multiple shaded sections to a plot, dependent on the value of a seperate time series data set (column of values). I believe I can use patch for this, but I have only seen it done where the time intervals are defined (hard coded), not dependent on a condition. The condition is simple: either positive or negative. See the attached image of what I'm trying to create - with shading going to ylims and plot area covered. Thanks

Answer 1

[Mathieu NOE]

hi

see demo code below - hope it helps

% dummy data
x = 0:99;
y = 1 + 0.02*sign(sin(1+x/25 + x.^2/250)) ; % y data 
y0 = 0.92 + zeros(size(y));  % create a y bottom line used latter on for the patch objects

y_gap = 0.01; % make a y gap between the main curve and the patch objects

cond1 = (y>1.01);
cond2 = (y<0.99);

% find start and end indexes of groups "cond1" and "cond2"
[begin1,ends1] = find_start_end_group(cond1);
[begin2,ends2] = find_start_end_group(cond2);

% loop over this groupes
figure(1)

for k = 1:numel(begin1)
    % groups "cond1"
    start1 = begin1(k);  % start index of k th groups of non zero values
    stop1 = ends1(k);  % end index of k th groups of non zero values
    ind1 = (start1:stop1);
    X=[x(ind1),fliplr(x(ind1))];                %#create continuous x value array for plotting
    Y=[y0(ind1),fliplr(y(ind1))-y_gap];              %#create y values for out and then back
    patch(X,Y,[0.9 0.9 0]);                  %#plot filled area
    hold on
end

for k = 1:numel(begin2)
    % groups "cond2"
    start2 = begin2(k);  % start index of k th groups of non zero values
    stop2 = ends2(k);  % end index of k th groups of non zero values
    ind2 = (start2:stop2);
    X=[x(ind2),fliplr(x(ind2))];                %#create continuous x value array for plotting
    Y=[y0(ind2),fliplr(y(ind2))-y_gap];              %#create y values for out and then back
    patch(X,Y,[0.5 0.6 0.9]);                  %#plot filled area
end

plot(x,y,'r','linewidth',2)
ylim([0.92 1.02]);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [begin,ends] = find_start_end_group(ind)
% This locates the beginning /ending points of data groups
% Important : ind must be a LOGICAL array

    D = diff([0;ind(:);0]);
    begin = find(D == 1);
    ends = find(D == -1) - 1;
end

[Numbers]

Hi Mathieu, thank you - this code has most of what I need. Is there a simpler way of finding the time span when each condition is true, rather than calling the [begins,ends] function? This function seems to add fine gaps of white space between my shaded sections, even when I changed the condition to be continuous - how do I remove the white space?

Also, some more info that was not clear in my question and the figure I shared, sorry: I don't need gaps between the shaded blocks and the plotted data - it can shade to the ylims. I've cleared out some of the unecessary code in the following:

clear; close all;

% dummy data

x = 0:99;

y = 1 + 0.02*sign(sin(1+x/25 + x.^2/250)) ; % y data

cond1 = (y>=1.0);

cond2 = (y<1.0);

% find start and end indexes of groups "cond1" and "cond2"

[begin1,ends1] = find_start_end_group(cond1);

[begin2,ends2] = find_start_end_group(cond2);

% loop over this groupes

figure(1)

plot(x,y,'r','linewidth',2)

for k = 1:numel(begin1)

start1 = begin1(k); % start index of k th groups of non zero values

stop1 = ends1(k); % end index of k th groups of non zero values

ind1 = (start1:stop1);

yl = ylim;

qy = [[1 1]*yl(1) [1 1]*yl(2)];

qx = [start1 stop1 stop1 start1];

patch(qx, qy, 'c')

hold on

end

for k = 1:numel(begin2)

start2 = begin2(k); % start index of k th groups of non zero values

stop2 = ends2(k); % end index of k th groups of non zero values

ind2 = (start2:stop2);

yl_1 = ylim;

ry = [[1 1]*yl_1(1) [1 1]*yl_1(2)];

rx = [start2 stop2 stop2 start2];

patch(rx, ry, 'y')

end

plot(x,y,'r','linewidth',2)

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

function [begin,ends] = find_start_end_group(ind)

% This locates the beginning /ending points of data groups

% Important : ind must be a LOGICAL array

D = diff([0;ind(:);0]);

begin = find(D == 1);

ends = find(D == -1) - 1;

end

[Mathieu NOE]

hello

you made a slight mistake in your code

the patch x coordinates are not simply the indexes but x(ind)

your code

qx = [start1 stop1 stop1 start1];

correct code

qx = [x(start1) x(stop1) x(stop1) x(start1)];

same for rx (see full code below)

so this modification avoid the x shift (dx = 1) between the red line and the patch objects as we can see in the picture you posted

now the remaining gap is blank because this is a transition aera, so I don't know what you want to do here

one solution to reduce that gap is to increase the x resolution

here with 100 samples (original code)

with 1000 samples (that seems good enough for my old eyes)

so if your data lacks some resolution , we can fix that by using interp1 and resample the data with finer resolution (my suggestion)

% dummy data 
N = 1000; % samples 
x = 0:N-1;
y = 1 + 0.02*sign(sin(1+2*pi*x/N + 0.01*x.^2/N)) ; % y data 
cond1 = (y>=1.0);
cond2 = (y<1.0);
% find start and end indexes of groups "cond1" and "cond2"
[begin1,ends1] = find_start_end_group(cond1);
[begin2,ends2] = find_start_end_group(cond2);
% loop over this groupes
figure(1)
plot(x,y,'r','linewidth',2)
for k = 1:numel(begin1)
    start1 = begin1(k);  % start index of k th groups of non zero values
    stop1 = ends1(k);  % end index of k th groups of non zero values
    ind1 = (start1:stop1);
   
    yl = ylim;
    qy = [[1 1]*yl(1) [1 1]*yl(2)];  
    qx = [x(start1) x(stop1)  x(stop1)  x(start1)];
    patch(qx, qy, 'c')
    
    hold on
end
for k = 1:numel(begin2)
    start2 = begin2(k);  % start index of k th groups of non zero values
    stop2 = ends2(k);  % end index of k th groups of non zero values
    ind2 = (start2:stop2);
    yl_1 = ylim;
    ry = [[1 1]*yl_1(1) [1 1]*yl_1(2)];
    rx = [x(start2) x(stop2)  x(stop2)  x(start2)];
    patch(rx, ry, 'y')
    
end
plot(x,y,'r','linewidth',2)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [begin,ends] = find_start_end_group(ind)
% This locates the beginning /ending points of data groups
% Important : ind must be a LOGICAL array
    D = diff([0;ind(:);0]);
    begin = find(D == 1);
    ends = find(D == -1) - 1;
end

now we can also force the patches to increase their width on both sides with dx = mean(diff(x)) but according to my second suggestion above , that doesn't seem to be really the best choice , also this is not compatible with data where there would be a "neutral" zone with blank patch

here demo with again N = 100 samples

% dummy data 
N = 100; % samples 
x = 0:N-1;
y = 1 + 0.02*sign(sin(1+2*pi*x/N + 0.01*x.^2/N)) ; % y data 
cond1 = (y>=1.0);
cond2 = (y<1.0);

dx  = mean(diff(x));

% find start and end indexes of groups "cond1" and "cond2"
[begin1,ends1] = find_start_end_group(cond1);
[begin2,ends2] = find_start_end_group(cond2);
% loop over this groupes
figure(1)
plot(x,y,'r','linewidth',2)
for k = 1:numel(begin1)
    start1 = begin1(k);  % start index of k th groups of non zero values
    stop1 = ends1(k);  % end index of k th groups of non zero values
    ind1 = (start1:stop1);
   
    yl = ylim;
    qy = [[1 1]*yl(1) [1 1]*yl(2)];  
%     qx = [x(start1) x(stop1)  x(stop1)  x(start1)];
    qx = [x(start1)-dx/2 x(stop1)+dx/2  x(stop1)+dx/2  x(start1)-dx/2];
    patch(qx, qy, 'c')
    
    hold on
end
for k = 1:numel(begin2)
    start2 = begin2(k);  % start index of k th groups of non zero values
    stop2 = ends2(k);  % end index of k th groups of non zero values
    ind2 = (start2:stop2);
    yl_1 = ylim;
    ry = [[1 1]*yl_1(1) [1 1]*yl_1(2)];
%     rx = [x(start2) x(stop2)  x(stop2)  x(start2)];
    rx = [x(start2)-dx/2 x(stop2)+dx/2  x(stop2)+dx/2  x(start2)-dx/2];
    patch(rx, ry, 'y')
    
end
plot(x,y,'r','linewidth',2)
xlim([min(x) max(x)]);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [begin,ends] = find_start_end_group(ind)
% This locates the beginning /ending points of data groups
% Important : ind must be a LOGICAL array
    D = diff([0;ind(:);0]);
    begin = find(D == 1);
    ends = find(D == -1) - 1;
end

[Mathieu NOE]

FYI , this is one other option

climanomaly - File Exchange - MATLAB Central (mathworks.com)

with this Fex submission, you can get this result - maybe you can modifiy the function to fill the entire y axis as yous wish

N = 100; % samples 
ref = 1;
x = 0:N-1;
y = ref + 0.02*sign(sin(1+2*pi*x/N + 0.01*x.^2/N)) ; % y data 
figure
[hlin,href,htop,hbot] = climanomaly(x,y,ref,'top','c','bottom','y',...
    'mainline','r-','refline','k--');
hlin.LineWidth = 2;
href.LineWidth = 1;
href.Visible  = 'off';
alpha(htop,0.7)
alpha(hbot,0.7)