Multivariate linear models and repeated measurements revisited

TY - GEN

T1 - Multivariate linear models and repeated measurements revisited

AU - Dalgaard, Peter

N1 - Conference code: 56

PY - 2009

Y1 - 2009

N2 - Methods for generalized analysis of variance based on multivariate normal theory have been known for many years. In a repeated measurements context, it is most often of interest to consider transformed responses, typically within-subject contrasts or averages. Efficiency considerations leads to sphericity assumptions, use of F tests and the Greenhouse-Geisser and Huynh-Feldt adjustments to compensate for deviations from sphericity. During a recent implementation of such methods in the R language, the general structure of such transformations was reconsidered, leading to a flexible specification method involving differences between orthogonal projections onto subspaces generated by within-subject models.

AB - Methods for generalized analysis of variance based on multivariate normal theory have been known for many years. In a repeated measurements context, it is most often of interest to consider transformed responses, typically within-subject contrasts or averages. Efficiency considerations leads to sphericity assumptions, use of F tests and the Greenhouse-Geisser and Huynh-Feldt adjustments to compensate for deviations from sphericity. During a recent implementation of such methods in the R language, the general structure of such transformations was reconsidered, leading to a flexible specification method involving differences between orthogonal projections onto subspaces generated by within-subject models.

M3 - Article in proceedings

SP - 3649

EP - 3652

BT - Bulletin of the International Statistical Institute

A2 - Gomes, Ivette

A2 - Silva, Pedro Luis do Nascimento

A2 - Alberto Silva, Jose

T2 - 56th Session of the ISI International Statistical Institute

Y2 - 22 August 2007 through 29 August 2007

ER -