2 REDUCED-BIAS AUTOCORRELATION ESTIMATORS - R(F1) AND R(F2)

Among the problems associated with the application of time-series anal ysis to typical psychological data are difficulties in parameter estim ation. For example, estimates of autocorrelation coefficients are know n to be biased in the small-sample case. Previous work by the present authors has shown that, in the case of conventional autocorrelation es timators of rho1, the degree of bias is more severe than is predicted by formulas that are based on large-sample theory. Two new autocorrela tion estimators, r(F1) and r(F2), were proposed; a Monte Carlo experim ent was carried out to evaluate the properties of these statistics. Th e results demonstrate that both estimators provide major reduction of bias. The average absolute bias of r(F2) is Somewhat smaller than that of r(F1) at all sample sizes, but both are far less biased than is th e conventional estimator found in most time-series software. The reduc tion in bias comes at the price of an increase in error variance. A co mparison of the properties of these estimators with those of other est imators suggested in 1991 shows advantages and disadvantages for each. It is recommended that the choice among autocorrelation estimators be based upon the nature of the application. The new estimator r(F2) is especially appropriate when pooling estimates from several samples.