Abstract
Diagnosis of the precision of nine methods for estimation of the first order autocorrelation in a SxO (3xq) design.The aim of this paper is to evaluate the efficiency of nine methods, two MLE and seven of moments, for estimation of the AR(1) dependence. A Monte Carlo simulation study was undertaken where two correlation structures, AR and ARH, were used. We manipulate the degrees of freedom in a SxO (3xq) design by means of various combinations of the sample size and the time points. The results indicate that when the structure is AR, the HCH, J, PP and WHS procedures estimate correcty the direction and the intensity of the autocorrelation. Otherwise, only the heterogeneity within groups increasing modify the PP and WHS estimation.
