Why Do MATLAB AR Coefficients Differ from Online Examples?

[verd]

Hi.

So I'm trying to do prediction/forecasting with a simple time series in MATLAB. But I'm having a bit of trouble with the AR coefficients.

So as a baseline, I've been trying to follow the example on this website:
http://local.wasp.uwa.edu.au/~pbourke/miscellaneous/ar/

In the first example, for the 3rd order example, they provide coefficients that look like:
2.753231 -2.740306 0.985501

But when I use MATLAB to generate the AR coefficients, using aryule or arburg, I get something completely different:
1.0000 -1.8242 0.9365

So when I loop something like:
x(t)=a(1)*x(t-1)+a(2)*x(t-2)+a(3)*x(t-3)
with the AR coefficients given to me at the above URL, I get a good prediction of the signal. But when I use MATLAB's coefficients (provided both from aryule and arburg) I get something that's not even close to the original signal.

So I was just wondering what I'm doing wrong? Am I using aryule/arburg incorrectly?

 

[KataKoniK]

Hi there,

It sounds like you are on the right track with using MATLAB's aryule and arburg functions to generate AR coefficients. However, it is important to note that different methods may produce slightly different coefficients due to differences in algorithms and implementation.

One possible explanation for the differences you are seeing is that the example on the website you mentioned may be using a different method or algorithm for generating the AR coefficients. It may be helpful to reach out to the author of the website or consult the documentation for their specific method to see if there are any differences.

Additionally, it is always a good idea to double check your code and make sure you are using the functions correctly. You can also try using different orders or methods for generating the AR coefficients to see if that affects the accuracy of your predictions.

I hope this helps and good luck with your time series forecasting!