code running for infinite while plotting my array for correlation

Hi,

I have problem as my code running for infinite time while I am plotting my arrays to achive coeffcient correlation.

here is my code:

for i=1:length(A)-300     
    Rx{i}= corrcoef(A (i:i+300), S(i:i+300));
    Time_Rx=i;
    B{1,i} = Rx(i)
    C{1,i} = Time_Rx
end

(Aiming for plot Rx as a function over time)

plotting by using:

plot(B,C);

Can not figure out what is wrong, Any help would be appriciable.

[Walter Roberson]

I see no evidence that it is running for infinite time. It is, though, doing a sliding window of correlation coefficients, sliding one further each time, and that operation could take some time. Are you sure that you want to slide only one each time?

Note that you are writing into cell arrays, but it is not permitted to plot() a cell array. Especially since it is a cell of cells -- you write into Rx{i} but you B{1,i} = Rx(i) and notice that Rx(i) is a 1 x 1 cell, not the result of the correlation coefficient but rather the cell wrapping the result of the correlation coefficient.

[Bob]

Thanks walter,

Attached snip from command window is the evidence for infinite run. it has been running since last more than an hour. I am not sure if I want only slide one each time. (How it should be implimented in either case).

If it is not permitted to plot() a cell array, what to do to achive my plot in this case?

B{1,i} = Rx(i)
C{1,i} = Rx

One small mistake it was

Rx{i}= corrcoef(A (i:i+300), S(i:i+300));

instead of

Rx{i}= corrcoef(A (i:i+300), B(i:i+300));

Kindly recommend some path for the plotting.

[Walter Roberson]

L = length(A) - 300;
B = zeros(2,2,L);
for K = 1 : L     
  B(:,:,K) = corrcoef(A(i:i+300), S(i:i+300));
end
C = 1 : L;

But it is not clear what you mean by plotting C (the index) vs B (a 2 x 2 array for each location).

It might make some sense to use

L = length(A) - 300;
B = zeros(L, 4);
for K = 1 : L
    cc = corrcoef(A(i:i+300), S(i:i+300));
    B(K, :) = cc(:);
end
C = 1 : L;
plot(C, B)
legend({'(1,1)', '(2,1)', '(1,2)', '(2,2)'})

[Bob]

Dear Walter,

There must be some confusion, I wanted to plot CC as correlation cofficent between the values A and S, over time.

Still its not achived.

[Walter Roberson]

When you calculate correlation coefficient between two columns, then the result is always 2 x 2.

Perhaps,

L = length(A) - 300;
B = zeros(L, 1);
for K = 1 : L
    cc = corrcoef(A(i:i+300), S(i:i+300))
    B(K) = cc(2,1);
end
plot(B);

[Bob]

Hi Again Walter,

When I use your inputs, following errors appeared:

Command window:

Warning: Colon operands must be real scalars.

Array indices must be positive integers or logical values.

cc = corrcoef(A(i:i+300), S(i:i+300))

[Walter Roberson]

L = length(A) - 300;
B = zeros(L, 1);
for i = 1 : L
    cc = corrcoef(A(i:i+300), S(i:i+300))
    B(i) = cc(2,1);
end
plot(B);

[Bob]

Wow. It worked well.

How can I make a Linear plot of CC values to visualize the all correlation points, either negative or positive correlation?

[Walter Roberson]

cc(1,1) is correlation of the section of A relative to itself. cc(2,2) is the correlation of the section of S relative to itself. cc(1,2) and cc(2,1) are the correlation of each relative to the other.

... I'm not sure what you are asking to extract and plot?

[Bob]

What I mean is, can I see eaither cc(1,2) or cc(2,1), as a linear correlation plot by using all the correlation data point from cc ?? some thing like attached figure (as example).

Also if I can plot one more figure

for Z : diffrence between A and S.

Z=A-S;
figure(2)
plot(B,Z);

or

plot(cc,Z);

Is it possible ? I tried and error appears in command window:

Error using plot

Vectors must be the same length.

[Walter Roberson]

p = polyfit(B,1);
hold on
t = 1 : length(A)-300;
plot(t, polyval(p, t), 'b-');

[Bob]

It shows error:

Error using polyfit (line 44)

X and Y vectors must be the same size.

p = polyfit(B,1);

[Walter Roberson]

t = 1 : length(A)-300;
p = polyfit(t, B, 1);
hold on
plot(t, polyval(p, t), 'b-');
hold off

[Bob]

still have the error:

Error using polyfit (line 44)

X and Y vectors must be the same size.

p = polyfit(t, B, 1);

[Walter Roberson]

p = polyfit(t(:), B, 1);

Example code attached.

[Bob]

strange plotting now, see attached Image.

If its not achivable its fine.

Can you help me plot another para meter in contineuation with same code:

A = randn(1e3,1); 
S = randn(1e3,1); 
L = length(A) - 300;
B = NaN(length(A), 1); 
for i = 1 : L
    cc = corrcoef(A(i:i+300), S(i:i+300)); 
    B(i) = cc(2,1);
end
plot(B);
Z=A-S;
figure(2)
plot(B,Z);

Can I plot B values at X axis and Z values at Y axis and in diffrent colors ?

[Walter Roberson]

A (nearly) horizontal line looks plausible to me given that data. Area above looks like it mostly balances area below, and the decreasing variance towards the right side would tend to anchor the endpoint of the line in values close to that range.

[Walter Roberson]

"Can I plot B values at X axis and Z values at Y axis and in diffrent colors ?"

When would the different colors be triggered?

Your B values increase and decrease irregularly like you see on the first plot. It would not seem to make sense to use them as the X axis values.

When you talk about "and in diffrent colors" my first interpretation was that you were wanting to plot two different lines, one for B, and one for Z, but then talking about plotting them on the X and Y axis would not seem to make any sense. The closest I can find to make sense of that would be to do something like

NA = length(A);
ZERO = zeros(NA, 1);
t = 1 : NA
plot3(t, B, ZERO, 'k', t, ZERO, Z, 'b');

which plots them with a common time base and with two different lines, each along a different axes.

[Jan]

[MOVED from flags] Bob wrote:

Worked well.

@Bob: Please use flags only to informa admins about inappropriate contents like rudeness or spam. Thanks.

[Bob]

Hi Walter,

How can I avraged my data (A anS) in minutes (for 10 minutes) before I do correlation calculation, in above code ??

Answer 1

[BERGHOUT Tarek]

if you want to plot Rx , then you should plot B not C, and you can't plot B vs C in this example because C and B they dont have the same length ( dimensions are not the same), try this code , I hope that it is helpful:

clear all;
clc;
%%%
t=1:600;
A=0.25;
B=1.25;
f1=0.5;
f2=0.025;
S=A*sin(f1*t)+B*cos(f2*t)+wgn(size(t,1),size(t,2),3);
A=S+randn(size(S));
%%%
B=[];
for i=1:length(A)-300     
    Rx=corrcoef(A (i:i+300), S(i:i+300));
    B=[B Rx];
end
plot(1:size(B,2),B);

[Bob]

Dear berghout,

How can I make a Linear plot of Rx values to visualize the all correlation points, either negative or positive correlation?