2nd Mplus Users Meeting
On Monday, March 14th the second meeting of the ‘Dutch Mplus user’s group: Exploring the boundaries of Mplus’ took place at Utrecht University, The Netherlands.
For a list of participants, click here.
The following presentations were given (including a link the the powerpoint presentations):
Key-note presentation (dr. Jan Boom)
Latent Growth Mixture Modelling
Smart applications of Growth Modeling in Mplus:
1) Evaluation of an intervention program in a longitudinal framework without the need for a control group (ppt).
2) Modeling the overlapping waves model of development as a linear growth model for a categorical outcome(ppt).
3)remaining issues, questions and answers (ppt).
An example of an Exploratory Structural Equation Model (dr. Mirjam van Zuiden) (ppt)
Exploratory Structural Equation Modeling (ESEM) with target rotation in longitudinal psychobiological research in the military.
Multiple group analysis in Mplus and Lavaan: (Yves Rossel) (ppt)
The presentation will focus on how to specify multi group models and how to test for measurement invariance? Furthermore, it will be shown what the defaults of Mplus are, and whether these defaults are useful or not (e.g. parameter constraints, meanstructure). The issue of defaults behind the scene of mutigroup analyses refers to a more general issue of Mplus: there are many constraints used in the background of the software, but how to recognize these defaults and how to override them. An alternative program is Lavaan where measurement invariance can easily be tested
Imputing Data: Why you should and how you could (Gerko Vink) (ppt)
What to do with missing data: FIML, single/multiple imputation.
Drifting apart or bonding together? Changes in parent-adult child relationships as the older generation ages into age old. An example of Latent Transition Analysis in Mplus (Martijn Hogerbrugge) (ppt)
Although the topic of intergenerational relations and support has been investigated previously,two major challenges remain unsolved. First, intergenerational relations have been investigated predominantly in a rather static way, mainly caused by the studies’ reliance on cross-sectional data. Second, previous studies have often selected only one or two dimensions of intergenerational relations in its research. Fortunately, latent transition analysis (LTA) offers a methodological solution for handling the complexity of intergenerational relations, while allowing the researcher to model transitions into different family types between time points. In this study, we will conduct LTA on panel data from the Longitudinal Study of Generations (LSOG), spanning a total of 17 years.
Seeking convergence (Peter Lugtig) (ppt)Computing the parameter estimates in a SE-Model in Mplus requires finding the the minimum of a loglikelihood function, that indicates the difference between the observed (S) and model implied (sigma) covariance matrices. MPLUS often has difficulties finding this minimum (or seeking convergence). This presentation will start with a short introduction in convergence algorithms. Then, we will discuss some situations in which convergence is difficult, and why this is the case. We will conclude with some suggestions for MPLUS model inputs that often ameliorate convergence problems.