In this work we deal with the measurement error issue arising in multilevel modelling when the sample cluster mean of a covariate is used in place of its population counterpart. We show that in a random intercept model with a single covariate the attenuation factor of the contextual effect is equal to the reliability of the covariate, so a simple correction is available. Some simulations confirm the analytical results and allow to study the sampling distribution of the corrected estimator under different hierarchical structures.
Measurement error in multilevel models with sample clustermeans / Grilli L.; Rampichini C.. - STAMPA. - Book of short papers:(2008), pp. 101-104. (Intervento presentato al convegno First joint meeting of the Société Francophone de Classification and the Classification and Data Analysis Group of the italian Statistical Society tenutosi a Caserta nel 11-13 Giugno 2008).
Measurement error in multilevel models with sample clustermeans
GRILLI, LEONARDO;RAMPICHINI, CARLA
2008
Abstract
In this work we deal with the measurement error issue arising in multilevel modelling when the sample cluster mean of a covariate is used in place of its population counterpart. We show that in a random intercept model with a single covariate the attenuation factor of the contextual effect is equal to the reliability of the covariate, so a simple correction is available. Some simulations confirm the analytical results and allow to study the sampling distribution of the corrected estimator under different hierarchical structures.
I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
