Paul Norris posted on Wednesday, January 03, 2007 - 8:30 am
Happy new year everybody. I have created a cross-sectional LCA model based on individuals’ responses to a large scale social survey (around 7000 respondents). I have now introduced several covariates to the model. These covariates are both at an individual level and at an area level so I am running the model using type = mixture missing twolevel with the appropriate variables included in within or between commands (all the “dependent” variables in the LCA are measured at the individual level, but covariates exist at both levels). The model runs successfully and I have been able to identify significant covariates from both the individual and area levels. Having identified significant covariates at both levels I was hoping to be able to make some statement about the relative importance of individual factors and area factors in explaining class membership. Does Mplus provide any means of assessing the relative importance of the different levels of covariates when conducting an LCA analysis (similar to those measures used in "normal" multilevel modelling)?
Many thanks in advance for any guidance anybody may be able to offer me with the above.
I don't know of any way to assess the relative importance of the different levels of covariates. The regression of the categorical latent class variables on the covariates answers the questions of which covariates predict class membership.
Paul Norris posted on Thursday, January 11, 2007 - 6:07 am
Many thanks for the clarification, much appriciated.