Multilevel mean differences?
Message/Author
 Melanie Fuchs posted on Friday, June 19, 2015 - 9:02 am
I want to calculate simple mean differences in a variable (y) depending on the belongniss to one of two institution types X (lower school, higher school). Of both types I have more than 15 institutions (cluster variable = inst). My question: How can I test for mean differences in achievement without neglecting the clustering? Is it right that I only need a between model?
BETWEEN = X;
CLUSTER = inst;
ANALYSIS: TYPE = TWOLEVEL;
MODEL:
%BETWEEN%
maa8 ON insttyp;
 Linda K. Muthen posted on Friday, June 19, 2015 - 4:44 pm

MODEL:
%WITHIN%
maa8;
%BETWEEN%
maa8 ON x;

The regression coefficient is the mean difference.
 Melanie Fuchs posted on Sunday, June 21, 2015 - 7:45 am
Thanks a lot! An additional question: How can it be that my regression coefficients (standardized) are very high, some are even higher than 1 (even though the mean differences in SPSS are quite small). How to interprete this?

MAA8 ON
INSTTYP 1.336 0.230 5.802 0.000
 Bengt O. Muthen posted on Sunday, June 21, 2015 - 6:40 pm
For us to be able to tell, you have to send the full output to support along with your license number.
 jen jang sheu posted on Saturday, September 05, 2015 - 6:22 am
I am now trying to estimate the student composition effect in an ANCOVA model. and the level 2 predictor will be the agreegated mean of the predictor in level 1 .How should I set the command in Mplus syntax and find the p value?

i.e. I want to find if Bc= r01-r10 is significant or not?

Thank you very much!
 Linda K. Muthen posted on Saturday, September 05, 2015 - 11:43 am
You can do that using MODEL TEST or MODEL CONSTRAINT. See the user's guide for further information on these commands.