The Missing Observations Theorem and a New Proof of Levinson’s Recursion
Charles W. Therrien · 2005
A principle of linear mean-square estimation which we call the "missing observations theorem" is presented here and used to provide a simple alternative derivation of Levinson's recursion. The derivation leads directly to the lattice form of the prediction error filter and avoids cumbersome matrix partitioning methods typically used in the derivation of the Levinson algorithm.