|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||||||
|
|
|
|
||||||||||||||||
|
|
||||||||||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||
|
|
|
|||||||||||||||||
|
|
ABSTRACT Recent Developments on Linear Mixed Models for Longitudinal Data, Possibly Subject to Dropout Geert Molenberghs,
Center for Statistics, Limburgs Universitaire Centrum; and
A general introduction to longitudinal data and the linear mixed model will
be presented. The topic will be approached from the modeller's and
practitioner's points of view. Apart from classical model building
strategies, many of which have been implemented in standard statistical
software, a number of flexible extensions and additional tools will be
discussed. Based on the linear mixed model setting, incomplete data
concepts will be reviewed and the issues arising with incomplete data will
be underscored. This will be done in both the selection as well as in
the
pattern-mixture frameworks. Then, it will be argued that an important way
forward is by means of sensitivity analysis. To this end, several tools
will be discussed, such as a local influence based approach and a method
based on so-called uncertainty regions. The use of pattern-mixture models
for sensitivity analysis will be considered. Finally, the impact of
incompleteness on efficiency will be assessed.
Return to Longitudinal/Survival Working Group List | Return to Home Page |
