5
Multilevel regression
Random intercept
Advantage
If people drop out you can still use data
Random intercept
every participant starts out at different level
Looks at BS differences
Identifiablility
Model is only meaningful when all components can be identified
nr of estimated parameter =/< nr of available estimates
If not identifiable
warning
Simplify model
Structures with random intercept added
ID+RI
only measures true correlation
participants are measured multiple times
V matrix
Spss combides G and R
CS
Same residual variance and covariance
AR1+RI
Special case of ARMA
Intra-class correlation
relliability
random intercept variance divided by measured variance
Factors
fixed
mean structure
interaction, linearity
Random
covariance
predictors
BS
WS
Difficult to destinguish in random part
intention to treat analyses
BV gender/treatment
SPSS
G matrix
BS variance
Only 1 element
R matrix
WS variance on diagonal
Most simple
Variance same for all timepoints
CS=RI ID