The effects of correlated errors on generalizability and dependability coefficients
This study investigated the effects of correlatederrors on the person x occasion design in which theconfounding effect of equal time intervals results incorrelated error terms in the linear model. Two specificerror correlation structures were examined: thefirst-order stationary autoregressive (SARI), and thefirst-order nonstationary autoregressive (NARI) withincreasing variance parameters. The effects of correlatederrors on the existing generalizability and dependabilitycoefficients were assessed by simulatingdata with known variances (six different combinationsof person, occasion, and error variances), occasionsizes, person sizes, correlation parameters, and increasingvariance parameters. Estimates derived from thesimulated data were compared to their true values. The traditional estimates were acceptable when the errorterms were not correlated and the error variances wereequal. The coefficients were underestimated when theerrors were uncorrelated with increasing error variances.However, when the errors were correlated with equalvanances the traditional formulas overestimated bothcoefficients. When the errors were correlated withincreasing variances, the traditional formulas both overestimatedand underestimated the coefficients. Finally,increasing the number of occasions sampled resulted inmore improved generalizability coefficient estimatesthan dependability coefficient estimates. Index terms:changing error variances, computer simulation, correlatederrors, dependability coefficients, generalizabilitycoefficients.
|Year of publication:||
|Authors:||Bost, James E.|
|Type of publication:||Article|
|Other identifiers:|| |
Saved in favorites
Similar items by subject
Find similar items by using search terms and synonyms from our Thesaurus for Economics (STW).