von Rosen, Dietrich
- Institutionen för energi och teknik, Sveriges lantbruksuniversitet
- Linköpings universitet
Sammanfattning
The aim of this work is to develop new methodology to detect influential observations in cross-over design models with random individual effects. Various case-weighted perturbations are performed. We derive the exact solution of influence of the perturbations on each parameter estimate and their dispersion matrix. Closed-form maximum likelihood estimates (MLEs) of variance parameters as well as fixed effect parameters in the cross-over design models are utilised. The work exhibits the possibility to produce closed-form expressions of the influence using the residuals in mixed models. A discussion on restrictions of the case-weighted perturbation schemes is given. Some graphical tools are also presented.
Nyckelord
cross-over design modelling; explicit maximum likelihood estimate; influential observation; mixed linear model; perturbation scheme; statistical diagnostics
Publicerad i
Research report - Department of Statistics, University of Stockholm
2011, nummer: 2011:1
Utgivare: Department of Statistics, Stockholm University
SLU författare
UKÄ forskningsämne
Sannolikhetsteori och statistik
Permanent länk till denna sida (URI)
https://res.slu.se/id/publ/37296