[R-sig-ME] LME4: interpretation of multivariate interaction effect
Eiko Fried
torvon at gmail.com
Tue May 8 18:52:46 CEST 2012
Dear Mailing List,
My model is this:
> m1<-lmer( Y ~ -1 + as.factor(Y_index) * X + (-1+as.factor(Y_Index)|subject), data=data2, REML=FALSE)
I have 9 response variables (items from a screening instrument), and
reorganized the data to have them in one row ("Y"). The variable
"Y_index" denominates the 9 different variables for Y (e.g. Y_index=2
is my second response variable).
The data are in a long format, every subjects has 9 lines.
I have a hard time interpreting the interaction output.
Fixed effects:
Estimate Std. Error t value
as.factor(Y_index)1 0.46177 0.05965 7.742
as.factor(Y_index)2 0.40207 0.05908 6.806
as.factor(Y_index)3 0.44255 0.07527 5.879
as.factor(Y_index)4 0.92783 0.07236 12.822
as.factor(Y_index)5 0.55113 0.07881 6.993
as.factor(Y_index)6 0.31863 0.06442 4.946
as.factor(Y_index)7 0.28208 0.06152 4.585
as.factor(Y_index)8 0.17036 0.04734 3.598
as.factor(Y_index)9 0.10635 0.03836 2.772
X 0.02473 0.03732 0.663
as.factor(Y_index)2:X 0.04296 0.03197 1.344
as.factor(Y_index)3:X 0.16419 0.04482 3.664
as.factor(Y_index)4:X 0.10612 0.04076 2.604
as.factor(Y_index)5:X 0.09108 0.04480 2.033
as.factor(Y_index)6:X 0.06488 0.03307 1.962
as.factor(Y_index)7:X 0.03331 0.03633 0.917
as.factor(Y_index)8:X -0.01012 0.03591 -0.282
as.factor(Y_index)9:X -0.05029 0.03893 -1.292
Do I understand correctly that "X" here is (Y_index)1:X, and the
reference line for the interpretation of the other interaction lines?
E.g. (Y_index)5, my 5th response variable, is [Estimate] .09 higher
than (Y_index)1, my first response variable, and with a t of 2.033
probably significantly higher? That is, X affects my fifth response
variable significantly stronger than my first response variable?
Thank you for your patience
E
More information about the R-sig-mixed-models
mailing list