|
|
Grandmean centering in type = twolevel |
|
Message/Author |
|
student07 posted on Thursday, August 02, 2007 - 1:23 am
|
|
|
Dear Drs Muthén I wonder how to use grandmean centering when using (1) type = twolevel and (2)type = twolevel random. Problem 1: Assume I have a model with both within- and between-level covariates. The model includes random intercepts only and is similar to Ex 9.3. I would like to conduct grandmean centering for the within- and between variables in order to make the intercept more interpretable. Q#1: Are there any general circumstance where grandmean centering would conflict with the tpye = twolevel approach? I came across this problem because I did not find (perhaps because its so late)any examples in the user guide where grandmean centering is applied for models using random intercepts only. Problem 2: Assume I have a model with both within- and between-level covariates, and the model includes random intercepts as well as random slopes, similar to Ex 9.1. I intend to use grandmean centering to mak the slope more interpretable. However it seems that in Ex. 9.1. centering is applied to the WITHIN-(x)variable only, but not for the BETWEEN-(w)variable. Q#2: Does grandmean centering of within- and between-level variables conflict with the assumptions underlying "type = twolevel random" ? P.S.: Is there any literature available which covers the statistical assumptions underlying "type = twolevel random" in the mplus framework? |
|
student07 posted on Thursday, August 02, 2007 - 2:37 am
|
|
|
May I ask another question: to request the between- and within covariance matrices in order to conduct preliminary analyses for a final twolevel factor analysis (step 3 and 4 in Muthén 1994), should one use the MEANSTRUCTURE command such as ANALYSIS: TYPE=MEANSTRUCTURE TWOLEVEL; and then SAVEDATA: SIGB IS Between.dat; !between covariance matrix SAMPLE IS Within.dat; !pooled within covariance matrix Thanks! |
|
|
Q#1: No. Q#2: No. P.S. See the Raudenbush and Bryk book: Raudenbush, S.W. & Bryk, A.S. (2002). Hierarchical linear models: Applications and data analysis methods. Second edition. Newbury Park, CA: Sage Publications. Yes, but you don't need MEANSTRUCTURE. It is the default with TWOLEVEL. |
|
Back to top |
|
|