How to plot the mean value of X corresponding to each Y value?

Hello!

I want to plot a vertical profile of temperature (so Y-axis is depth (m) and X-axis is temperature (ºC)). Because I have many measurements for each depth, what I get is a profile made of a cloud of points:

However, I would like to get a linear profile, so I guess I would need to plot each X mean value for each Y, right?

Could someone please help me on that?

Thanks a lot!

[Mathieu NOE]

hello

please provide your data (attach a mat file)

do you wnt the mean line or the upper / lower envelope ?

[Alejandra Uguet de Resayre]

It doesn't let me attach the mat file because it is too big (more than 5MB).

What I mean is that from that cloud of points I want to get something like this instead, like a single line thing, so I guess that's the mean value of X for each Y.

Thank you for answering!! <3 <3 <3

[Star Strider]

@Alejandra Uguet de Resayre — Please see my latest Comment to my Answer. It produces the result matching the plot you posted.

Answer 1

[Star Strider]

I am not certain what you want.

Perhaps something like this —

x = ones(9,1)*linspace(13, 19, 50);                                 % Create Data
y = 8E+2*(1-exp(-5*(x-13)))-1E+3 + randn(9,size(x,2))*75;           % Create Data
T = x(:);                                                           % Temperature (Assume Column Vector)
D = y(:);                                                           % Depth (Assume Column Vector)

NBins = 14;                                                         % Change This To Change Temperature Resolution
[N,Edges,Dpthv] = histcounts(D, NBins);

Ctrs = Edges(1:end-1) + mean(diff(Edges));                          % Depth Centers
% Tvct = linspace(min(T), max(T), NBins);
TempMean = accumarray(Dpthv, T, [], @mean)                          % Calculate Mean Temperature At Each Depth

TempMean = 14×1

13.0000 13.0000 13.0000 13.0000 13.1224 13.1224 13.1224 13.1837 13.6234 15.5923

figure
plot(T, D, '.r', 'DisplayName','Data')
hold on
plot(TempMean, Ctrs, '.-b', 'DisplayName','Mean Temperature')
hold off
grid
xlabel('Temperature (°C)')
ylabel ('Depth (m)')
legend('Location','best')

The blue line plots depth as a function of the mean temperature.

.

[Alejandra Uguet de Resayre]

yes it is that! but that code is not working for me, I don't know why :(

[Star Strider]

Thank you!

Please upload your data. I will do my best to get my code to work with it.

[Alejandra Uguet de Resayre]

This is it!

theta is the potential temperature and dep is the depth :)

[Star Strider]

The only significant adjustment to my code to make it work with your data is to add 'omitnan' to the mean call in the accumarray call —

TempMean = accumarray(Dpthv, T, [], @(x)mean(x,'omitnan'))

I assigned ‘theta’ to ‘T’ and ‘dep’ to ‘D’ and increased ‘NBins’ to 50 to provide a smoother curve, and negated ‘D’ and‘Ctrs’ in their respective plot calls.

Try this —

LD = load(websave('matlab','https://www.mathworks.com/matlabcentral/answers/uploaded_files/1259980/matlab.mat'));
candec18tCopy = LD.candec18tCopy;
theta = candec18tCopy.theta;
dep = candec18tCopy.dep;

T = theta;                                                          % Temperature (Assume Column Vector)
D = dep;                                                            % Depth (Assume Column Vector)
NBins = 50;                                                         % Change This To Change Temperature Resolution
[N,Edges,Dpthv] = histcounts(D, NBins);
Ctrs = Edges(1:end-1) + mean(diff(Edges));                          % Depth Centers
Tvct = linspace(min(T), max(T), NBins);
TempMean = accumarray(Dpthv, T, [], @(x)mean(x,'omitnan'))          % Calculate Mean Temperature At Each Depth

TempMean = 50×1

15.9333 15.9292 15.8779 15.7944 15.5730 15.3716 14.8180 14.3202 14.0491 13.8852

figure
plot(T, -D, '.r', 'DisplayName','Data')
hold on
plot(TempMean, -Ctrs, '.-b', 'DisplayName','Mean Temperature')
hold off
grid
xlabel('Temperature (°C)')
ylabel ('Depth (m)')
legend('Location','best')

EDIT — Minor format improvements. Code unchanged.

.

[Alejandra Uguet de Resayre]

This one also worked perfectly :-) Is there anyway I can accept both answers? I don't know which one to choose! Thank you so so much for your time

[Star Strider]

You originally accepted mine (this one) because it produced the result you said you wanted.

It still does, specifically (calculating and) ‘plotting the mean value of X corresponding to each Y value’. I was off doing other things (I have a life outside of MATLAB Answers) and did not immediately see that you had attached your data. After I saw it, the reason you could not get my code to work with it (the NaN values) became immediately obvious, and I made a minor change in the ‘TempMean’ calculation to work with it. With that change it ran with your data without error.

[Mathieu NOE]

dear all

I think that after all @Star Strider had a better solution so it's natural he wins the prize !

i probably interpreted the requirements in a wrong way and if for any reason my solution has still some value in the eyes of the OP then I am happy

have a good day

[Alejandra Uguet de Resayre]

You all are right :) Thank you both for your time invested, it helped me so much.

[Star Strider]

As always, my pleasure!

Answer 2

[Mathieu NOE]

hello

yes i'm coming late in the show... nevertheles I am here !!

just for the fun and FYI

your data could be decimated up to factor 100 without loss of quality I think

without decimation :

with decimation factor 50

code

load('matlab.mat');

temp = candec18tCopy.theta;
depth = -candec18tCopy.dep;

% remove nan's
idx = isnan(temp);
temp(idx) = [];
depth(idx) = [];
% sort 
[temp,id] = sort(temp,'ascend');
depth = depth(id);
% unique
[temp,IA,IC] = unique(temp);
depth = depth(IA);

% decimation
decim_fact = 50; % test with 1 to 100
ind = 1:decim_fact:numel(depth);
temp = temp(ind);
depth = depth(ind);

% remove data below 13.25
ind = (temp<13.25);
temp(ind) = [];
depth(ind) = [];


depth_s = smoothdata(depth,'movmedian',round(10000/decim_fact));
plot(temp,depth,'*r',temp,depth_s,'-b')

[Alejandra Uguet de Resayre]

yes yes yes!! thank you soooo much Mathieu

[Mathieu NOE]

My pleasure !

Glad it worked