How to know which Signal is delayed !

Dear all, I have my code below for real data samples. I already tested the xcorr delay calculation method with different data and it produced the right answer but with the real data, I am not sure how to know after calculating the delay which of the two signal is the delayed version of the other?

x = sample1(:,1);
X = (x).';
y = sample2(:,1);
Y = (y).';
figure;
clf
subplot(3,1,1);
[xi,f]=ksdensity(X); 
plot(f,xi); 
line(repmat(X,2,1),repmat([0;0.1*max(xi)],1,length(X)),'color','r' ); 
subplot(3,1,2);
[xi,f]=ksdensity(Y); 
plot(f,xi); 
line(repmat(Y,2,1),repmat([0;0.1*max(xi)],1,length(Y)),'color','r' );
[Rxx,lags] = xcorr(X,Y);
[Z,delay] = max(Rxx);
lags(delay);

Answer 1

[Fangjun Jiang]

Susan, Others don't have your data sample1 and sample2 so won't be able to look at your data. Would the sign of lags(delay), be positive or negative, tell you whether Y is the delay of X, or X is the delay of Y?

In my answer to your previous question, http://www.mathworks.com/matlabcentral/answers/13110-signal-processing

y is 10 steps lag behind x, when running

[Rxx, lags] = xcorr(x, y);
[Y, I] = max(Rxx);
lags(I)

It returns -10, meaning y is 10 steps lag behind x

If you run

[Rxx, lags] = xcorr(y, x);%switch order of x and y
[Y, I] = max(Rxx);
lags(I)

It returns 10, meaning x is 10 steps ahead of y

[Susan]

Yeah I thought the sign should tell me but the fact that for all my sample data I had a large positive number made me not very confident of the way I've done things, however the code works with artificial data so I suppose the methods works. Is there a chance that even though the methods works with artificial data(data we know the delay) it wouldn't work for a real data or it wont be efficient for it?

[Susan]

I actually tried to swap x,y so to check If I will get negative answer but I got a complete different answer??

[Fangjun Jiang]

You could visually inspect your data to get a sense whether the result is in the ballpark.

[Susan]

It would not be accurate but the fact that swapping both x,y gives different results I think its a sign the delay calculated is not right.. Any idea where I am going wrong?