Theoretical analysis derives approximate sampling distributions for circular serial correlation coefficients, indicating consistent parameter estimation in correlated data.
It is desired to find an approximate distribution of simple form for the statistic r = x₁x₂ + ⋯ + xTx₁/x₁² + ⋯ + xT² (r is an estimate of the serial correlation coefficient ρ in a circular universe) in the case that ρ ≠ O in the universe. Such a distribution is obtained by smoothing the joint characteristic function of the numerator and denominator of the expression for r. The first two moments are calculated; from these r is seen to be a consistent estimate of ρ. A graph of this distribution for sample size $T = 20$ and various values of ρ is given. In addition, an approximate distribution for p = x²₁ + ⋯ + x²T is derived which reduces to the exact (χ²-) distribution if ρ = 0. From a formula which yields all moments, it is concluded that, at least up to the degree of approximation attained, $p/T$ is an unbiased and consistent extimate of σ².
No takes yet. Share an insight, caveat, or question.
Roy B. Leipnik (1947) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: