For the general case AR(*p*), we can multiply the equation by *X*_{t-k}
and take the expectation value:

where is the autocovariance function at lag *k*.

Since from (7), we have

(9) |

(10) |

and therefore the estimates of AR coefficients can be solved.

Multiply (8) by *X*_{t},

**Maximum Entropy Method** (**MEM**, or **Burg algorithm**) is an alternative way to estimate AR coefficients.