

Fixed effects model in SEM 

Message/Author 

daj posted on Tuesday, February 23, 2010  4:42 pm



Hi, I have a survey dataset, in which each respondent evaluates two agents. Each evaluation is treated as an individual observation in my dataset (so I have n x 2 observations, where n is the number of respondents). I'm treating each respondent as a cluster. I'm trying to eliminate unobserved individual effects using the Fixed Effects method that is common in econometrics analysis  where we demean each observation with the cluster's average. My question is how can I do this in MPlus? There are two approaches that I can think of: 1. Set up a multilevel analysis. But I don't really have clusterlevel data (as we are focusing on the respondentagent relationship), and does this help to eliminate unobserved individual effects? 2. Do a groupmean centering for all variables, and demean each observation with its groupmean. Run the analysis using the demean variables. Does any of these sound right? Or is there a better way? Thanks! 


With only 2 agents within respondent it is like having twin data where seldom a multilevel approach is used but instead a multivariate approach. So you have n observations and you have 2 DVs (eval of agent 1 and eval of agent 2). I don't know about the demeaning approach. A paper talking about Fixed effects approaches and SEM is: http://papers.ccpr.ucla.edu/papers/PWPCCPR2008003/PWPCCPR2008003.pdf 

Back to top 

