Small sample estimation of ρ1

Hydrologists frequently need to estimate lag 1 correlation ρ 1 from rather short observed records of length n . The most commonly used algorithms for estimating ρ 1 can give seriously biased estimates if n is small and |ρ 1 | is large. Small sample bias corrections based on the assumption of a lag 1 Markov generating process are available for , but the corrected estimates tend to have larger variance than the uncorrected estimates do.