StatsWAP2009July17

<meta name="title" content="July 17, 2009"/>

Mixed Effects Models in MRI Research

Summer Seminar Series: Statistics without the Agonizing Pain (WAP Series)
Time/Location: Friday, July 17, 2009 – 3 pm – 4 pm Clark Hall 314
Speaker: Yang An, National Institute on Aging - Intramural Research Program

Laboratory of Personality and Cognition - http://www.grc.nia.nih.gov/branches/lpc/lpc.htmv
Cognition section - http://www.grc.nia.nih.gov/branches/lpc/sresnick.htm

Mixed effects models (continuous outcomes) - also known as random effects models
Generalized estimating equations (discontinuous outcomes)

General topics

• Ordinary regression model - each data point comes from a separate "person" - L2 fit - Fixed effects
• ME model - each subject has its own slope. Find population slope.
• REML - restricted maximum likelihood -- preferred for model estimation. Not suitable for model comparison with differing random effects.
• Empirical Bayes (EB) for estimation of individual trajectories

Links about R

• Fitting Linear Mixed Models in R : http://cran.osmirror.nl/doc/Rnews/Rnews_2005-1.pdf#page=27