,
11 tweets,
3 min read
Read on Twitter
Latent Class / Mixture models have been around my feed a lot recently. So I wanted to share with you all a little program that we* built to make running LCA-based models in Mplus less of a chore.
*We = Me, my non-academic friend @rihjol who developed the initial programming, and grad student @JamesUanhoro who added cool features and moved it online off of my desktop and into cyberspace so anyone could use it.
If you've ever run an LCA in Mplus you know that it's... a slow process.
You specify a model estimating two classes or profiles, run it, harvest a bunch of information from the output, then write code for a parallel three-class model, run it, harvest, four-class..... and so on
You specify a model estimating two classes or profiles, run it, harvest a bunch of information from the output, then write code for a parallel three-class model, run it, harvest, four-class..... and so on
That process of specifying, running, and harvesting was taking ages, and sometimes errors happen. Maybe in the final model, you meant to have a correlated error covariance but *WHOOPS* somehow you left that out of the three-class model. Now start over again.
Enter our Mplus LCA Helper!
The Helper works with your copy of Mplus. It will AUTOMATICALLY run multiple models for you, varying the number of classes, and saving the input and output files.
The Helper works with your copy of Mplus. It will AUTOMATICALLY run multiple models for you, varying the number of classes, and saving the input and output files.
It ALSO pulls critical information about model fit into one .csv file for you to make it easy to quickly compare which model fits best. That file looks like this: 
But wait, there's more.
That .csv file - below the model fit indices - also gives you the mean values (or logits if you're fitting an LCA) for each variable in each identified class. Excerpt here:
That .csv file - below the model fit indices - also gives you the mean values (or logits if you're fitting an LCA) for each variable in each identified class. Excerpt here:
How to use it:
It runs off of your own copy of Mplus, but you don't need to have it open. You copy and paste Mplus code for the two-class model into our little text box.
Use the pull-down boxes to select your operating system and the max number of classes you want it to fit.
It runs off of your own copy of Mplus, but you don't need to have it open. You copy and paste Mplus code for the two-class model into our little text box.
Use the pull-down boxes to select your operating system and the max number of classes you want it to fit.
There's a little button at the bottom "Download Mplus Automation File".
You put it in the same folder where your data is located.
Double click it.
And watch it go.
You put it in the same folder where your data is located.
Double click it.
And watch it go.
Got a little video of the whole process:
Want to try it out? Visit it here:
mplus-output-scraper.herokuapp.com
Love to hear feedback from you on how it's working!
mplus-output-scraper.herokuapp.com
Love to hear feedback from you on how it's working!
