Message/Author 


Does it make any sense and is it even possible to examine the impact of a level1 independent variable (e.g. student gender) on a level2 dependent variable (e.g. mean performance), which is an aggregated variable. When I try this with Analysis: type = TWOLEVEL, the output state: THE VARIANCE OF 'mean perfomance' APPROACHES 0. FIX THIS VARIANCE AND THE CORRESPONDING COVARIANCES TO 0, DECREASE THE MINIMUM VARIANCE, OR SPECIFY THE VARIABLE AS A BETWEEN VARIABLE What should I do? 


Any variable that does not vary within clusters must be put on the BETWEEN list. These variables can be used only in the between part of the model. A variable measured on the individual level can be used in both parts of the model if it is not placed on the WITHIN list. In this way you could regress a between variable on the between part of a within variable. See Examples 9.1 and 9.2 for further information. 

Jana Nie posted on Wednesday, June 08, 2011  3:47 am



Dear Ms Muthen, is it possible to use a level 2 dependent variable if I first aggregate (mean) the data on that level two variable? 


Yes. But also try to use the latent variable (random intercept) version of this Level 2 DV. 


On the other hand, is it possible to let the amount of variance at the level1 vary as a function of a level2 variable? 


If the level 2 variable is categorical, you could use it as a grouping variable in a multiple group analysis. 


I understand that multilevel regression cannot accommodate level 2 outcomes. Without thinking, I ran the following model and am now wondering why it worked. I have not modeled any latent variables. Therefore this is multilevel regression with a level 2 outcome and not structural regression? What am I missing? BETWEEN= GCP; DEFINE: GCP = MEAN(GCP1r GCP2r GCP3 GCP4 GCP6); ECP = MEAN(ECP1r ECP2r); ANALYSIS: TYPE= TWOLEVEL; ESTIMATOR= MLR; MODEL: %WITHIN% ECP; %BETWEEN% GCP ON ECP; MC 


That limitation is true for regular multilevel software, not for multilevel modeling in Mplus. See e.g. UG ex9.4. 

Back to top 