Message/Author 

Arena posted on Thursday, August 16, 2012  9:33 pm



I'm currently analyzing data to examine how classroom characteristics relate to students' standardized test scores. Because students are nested within classrooms, I used CLUSTER to adjust the standard errors within each classroom. However, I also want to employ a school fixed effects model to account for the unobserved characteristics of schools that may bias observed relationships. How can I do this in Mplus? Thank you in advance. Cheers, Arena 


I'm not sure what you mean. 

Arena posted on Friday, August 17, 2012  1:43 pm



I am wondering if there's a procedure in Mplus that is equivalent to the xtregfe command in Stata. 


It sounds like this is a multilevel model where students are nested in classrooms and dummy variables are used as covariates to represent schools. Yes, this can be done in Mplus. 


I have the same question above. Would you please be more specific on the ways in which school dummies could be specified in the model.I have students (n=20,000) nested within schools (n=89). Since I was not interested in estimating group level variations, I had used the "cluster" option to take into account a clustering issue. Now reviewers want me to use fixed effects dummies for schoolto make sure that any variance resulting from the group id could be adequately partialled out, instead of the cluster option. Thank you. 


School dummy variables are used as covariates. I would not recommend this in your case with 88 dummy variables. I would instead use TYPE=TWOLEVEL. The results you want would be on the within level. You could have a model of only random intercepts on the between level. See Example 9.1. 

Markus Riek posted on Thursday, January 16, 2014  12:48 pm



Fixed effects in a complex (crossnational) sample. I’m doing a SEM analysis based on a sample of multiple countries. I want to analyze the effects for the whole sample (not between the countries). In order to account for fixed effects I specified the following parameters in the analysis configuration. VARIABLE: WEIGHT = w; CLUSTER = country; ANALYSIS: TYPE = COMPLEX; Is that the correct specification to include country fixed effects? Does someone know about an example of a singlelevel analysis including fixed effects? Thanks, Markus 


"Country fixed effects" sounds to me like a multiplegroup analysis where group is country (a singlelevel analysis). The setup you give would not estimate any such effects  it would just correct the SEs for country clustering. A related paper discussing fixed versus random effects is on our website: Muthén and Asparouhov (2013). New methods for the study of measurement invariance with many groups. Mplus scripts are available here. 

shonnslc posted on Monday, May 13, 2019  7:34 pm



Hi, I am conducting path analysis with multilevel data (cluster = 6). I am aware that with this small number of clusters, type = complex + MLR is not a viable solution. Therefore, I am trying fixed effects model instead with five dummy variables as covariates. The output showed that the model estimation terminated normally. However, a warning showed up (ec is one of the endogenous variables; q4_3 is one of the dummy variables): THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NONPOSITIVE DEFINITE FIRSTORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS 0.856D16. PROBLEM INVOLVING THE FOLLOWING PARAMETER:Parameter 21, EC ON Q4_3 In this case, can I still trust my results? Thanks. 


You want to find the reason for this message. For instance, if you brought the dummy vble into the model, the message is harmless. 

Back to top 