Calculate Exponetial Moving Covariance

I should calculate a variance-covariance matrix over a 60-month rolling window, but so that the newer data has more weight than the older data, such as an exponential moving average. Can you give me a hand? thank you in advance

Answer 1

[Jeff Miller]

Not really a hand, but maybe some hints...Let's say you have variables x and y, and you want the exponentially weighted rolling means, variances, and covariances.

It looks like you can get the means with dsp.MovingAverage (though I haven't done this so am not sure of the details).

To get the variances and covariances, form the new variables x.^2, y.^2, and x.*y. You can also compute the moving averages of these new variables with dsp.MovingAverage.

Finally, at each time point compute the variance as e.g., [rolling average of x.^2 - (rolling average of x)^2] and the covariance as [rolling average of x.*y - (rolling average of x)*(rolling average of y)]