Correlation coefficient calculation error

As I try to find correlation coefficient for the respective variables i.e. 'td' & 'cumulat', in for loop, some error is coming in the process of execution. Please help.

clear;clc;close all
S = load('DQS.mat');
F = fieldnames(S);
for n = 1:numel(F)
    F1 = fieldnames(S.(F{n}));
    F11{n} = F1;
    % allocate the variables
    nn = cell(numel(F1),1);
    PArea = cell(numel(F1),1);
    t_max = cell(numel(F1),1);
    for n1 = 1:numel(F1)
        t = S.(F{n}).(F1{n1}).cc.t;
        t_max{n1} = max(t);
        nn{n1} = F1{n1};
        output{n1} = cumtrapz(t);
        NF = @(p,q) max(output{n1}(t<=q)) - min(output{n1}(t>=p));
        PArea{n1} = NF(5, 6);   
    end  
    T_max{n} = t_max;
    tt = cell2mat(T_max{1,n}(:,end));
    td{n} = tt./max(tt);
    cumulat{n} = PArea;
    [R{n},P{n}] = corrcoef(td,cumulat); %correlation coeff. calculation
end

Error using corrcoef
Cannot compute p-values for complex inputs.

[Jonas]

can you please edit your question, format your code as code and press the compile button (green rectangle), this way we can see what is happening

Answer 1

[Mathieu NOE]

hello

seems to me the code could be made simpler, why the need to make all this cell acrobatics ?

clear;clc;close all
S = load('DQS.mat');
F = fieldnames(S);
for n = 1:numel(F)
    F1 = fieldnames(S.(F{n}));
    F11{n} = F1;
    % allocate the variables
    nn = cell(numel(F1),1);
    for n1 = 1:numel(F1)
        t = S.(F{n}).(F1{n1}).cc.t;
        t_max(n1) = max(t);
        nn{n1} = F1{n1};
        output{n1} = cumtrapz(t);
        NF = @(p,q) max(output{n1}(t<=q)) - min(output{n1}(t>=p));
        PArea(n1) = NF(5, 6);   
    end  
td = t_max./max(t_max);
cumulat = PArea;
[R{n},P{n}] = corrcoef(td,cumulat); %correlation coeff. calculation
end

[Josh]

DQS structure array has two fields: 'xc1' & 'xc2'. 'xc1' has three fields-'c1','c2', & 'c3' whereas xc2 has two fields-'c1', and 'c2'. Now, the output of the modified code is for field 'xc2' with the thrird row of field 'xc1' as it is not overwritten. The cell operations in the previous code was utilized to keep both fields i.e. 'xc1' & 'xc2'. Therefore this modified code is not resolving the earlier issue. Please reconsider.

[Mathieu NOE]

In your original code you use corrcoef with cell arrays as inputs

that does not work, corrcoef needs numerical array

clear;clc;close all
S = load('DQS.mat');
F = fieldnames(S);
for n = 1:numel(F)
    F1 = fieldnames(S.(F{n}));
    F11{n} = F1;
    % allocate the variables
    nn = cell(numel(F1),1);
    PArea = cell(numel(F1),1);
    t_max = cell(numel(F1),1);
    for n1 = 1:numel(F1)
        t = S.(F{n}).(F1{n1}).cc.t;
        t_max{n1} = max(t);
        nn{n1} = F1{n1};
        output{n1} = cumtrapz(t);
        NF = @(p,q) max(output{n1}(t<=q)) - min(output{n1}(t>=p));
        PArea{n1} = NF(5, 6);   
    end  
    T_max{n} = t_max;
    tt = cell2mat(T_max{1,n}(:,end));
    td{n} = tt./max(tt);
    cumulat = PArea;
    [R{n},P{n}] = corrcoef(td{n},cell2mat(cumulat)); %correlation coeff. calculation
end

[Mathieu NOE]

In terms of results R{n},P{n} both suggestions gives the same results

[Josh]

As the outputs are mostly 'NaN' values, it seems the inputs to 'corrcoef' function are not the numerical values of variables i.e. 'td' & 'cumulat'. Please reconsider & see if it can be resolved.

[Mathieu NOE]

they are numerical

first iteration (n= 1)

both inputs are numerical :

K>> td{n} =

0.8333

0.7500

1.0000

K>> cell2mat(cumulat) =

5.5000

5.5000

5.5000

result is :

corrcoef(td{n},cell2mat(cumulat)) =

1 NaN

NaN NaN

[Mathieu NOE]

and for the second iteration , same comments

K>> td{n} =

0.7500

1.0000

K>> cell2mat(cumulat) =

5.5000

5.5000

K>> corrcoef(td{n},cell2mat(cumulat)) =

1 NaN

NaN NaN

[Josh]

Thanks for your kind reconsiderations.It is resolved.

[Mathieu NOE]

OK

all the best