Efficiency of the Sample Mean when Residuals Follow a First-Order Stationary Markoff Process

The efficiency of the sample mean relative to the “best” linear unbiased estimator when residuals follow a first-order stationary Markoff process is studied. The minimum efficiency for correlation ρ ≧ 0 occurs at the unique root of a certain equation depending on the sample size n and decreases monotonically in n toward the value .878 obtained by Ylvisaker in the continuous case. It is tabulated for selected n ≦ 300, together with the value of p at which the minimum is achieved.