difficulty interpreting outputs of movcorr

im using movcorr from the file exchange to calculate moving correlation between two time series.

I have 2 time series, interpolated onto the same dt, which contain 1534 datapoints. I have used movcorr to calculate a moving correlation with a 100 point window. When I do this, both r and p contain 1534 points, so surely the window is not being applied correctly as I expected an output of 1 correlation coefficient every 100 points?

I have tried to look at the documentation on the file exchange but I am having difficulty intepreting this output.

k = 100;
[r, p] = movcorr(data_1_interp,data_2_interp,k);

Answer 1

[Ronit]

Hello CG,

The function movcorr calculates the correlation coefficient for each point in the time series using a sliding window of specified length (in your case, 100 points). This means that each point in the output arrays r and p represents the correlation coefficient calculated over a window centred at that point, hence why you get a result for each of the 1534 points.

If you expected an output of one correlation coefficient every 100 points, you might consider modifying your approach. Here a suggestion:

After computing r and p, you can take every 100th value to get a reduced set of correlation coefficients.

downsampled_r = r(1:100:end);
downsampled_p = p(1:100:end);

I hope it helps with your query!

[LC]

Thank you for your response, that makes sense. I had assumed this function worked like movsum.