The Effects of Correlated Errors on Generalizability and Dependability Coefficients

This study investigated the effects of correlated on the person x occasion design in which the effect of equal time intervals results in error terms in the linear model. Two specific correlation structures were examined: the -order stationary autoregressive (SARI), and the -order nonstationary autoregressive (NARI) with variance parameters. The effects of correlated on the existing generalizability and dependability were assessed by simulating with known variances (six different combinations person, occasion, and error variances), occasion , person sizes, correlation parameters, and increasing parameters. Estimates derived from the data were compared to their true values. The traditional estimates were acceptable when the error were not correlated and the error variances were . The coefficients were underestimated when the were uncorrelated with increasing error variances. , when the errors were correlated with equal the traditional formulas overestimated both . When the errors were correlated with variances, the traditional formulas both overestimated underestimated the coefficients. Finally, the number of occasions sampled resulted in improved generalizability coefficient estimates dependability coefficient estimates. Index terms: error variances, computer simulation, correlated , dependability coefficients, generalizability .