Mplus 8.0 with Dynamic Structural Equation Models

Most of you know Mplus, a statistical software that is considered by many to be state-of-the-art for Structural Equation Models (SEM).

The network literature on psychopathology so far has been dominated by R, for instance when estimating cross-sectional network models via the packages qgraph, IsingFit, mgm, or bootnet, or time-series network models via mlVAR or graphicalVAR.

Modeling intensive time-series processes has become a topic important enough that the Mplus team asked Ellen Hamaker from Utrecht University to collaborate regarding the implementation of such models. I don’t know the details, but an invitation to a workshop states that Mplus will be able to estimate dynamical processes in data collected via “daily diaries, ecological momentary assessments (EMA), experience sampling methodology (ESM), and ambulatory assessments”. As usually, the Muthéns have found a sweet name and abbreviation for this: Dynamic Structural Equation Models (DSEM).

If you are interested in learning more about Mplus 8 that will come out this spring, there are several workshops planned. The first will take place July 13/14 2017 at Utrecht University, and I believe there will also be presentations by early adaptors of DSEM. If you are using Mplus 8 and want to present, email Rens van de Schoot.

Personally, I “grew up” with Mplus and it brings great benefits like fantastic state-of-the-art SEM modeling and very good support. Over the years, I have slowly transitioned to R because it is free open source software, deals much better with data management (with Mplus I usually need a second program to do the data management), and allows me to make my analyses 100% reproducible (everybody can get R and run my scripts, while not everybody can purchase Mplus and run my scripts). No matter how you stand on this debate, we should all be excited that Mplus is jumping on the dynamical process modeling bandwagon: this guarantees the thorough implementation of novel methodology, and the more software is out there that supports and further develops these models the better.

Leave a Reply

Your email address will not be published. Required fields are marked *