I am trying to design a predictive AR filter using aryule. The code looks like this:
h = aryule(signal,order);
est_signal1 = filter(1,h,signal);
est_signal2 = conv(-h(2:end),signal);
When I do this, the est_signal1 has a different amplitude than the original (generally larger). However, est_signal2 is much more similar (so long as you cut off the final 'order' number of entries). But the AR model is an all pole filter, so using filter(1,h,signal) should work the same as conv(-h(2:end),signal), right?