Installing the development version of lavaan

If you have not installed lavaan before, first install the current version of lavaan and all of its dependencies. Then, to install the latest development version of lavaan, you can type at the R prompt:

library(remotes)
remotes::install_github("yrosseel/lavaan")

To make sure you are using the newly installed version of lavaan, restart your R session. If you want to revert to the official (CRAN) version of lavaan again, simply type

install.packages("lavaan")

and you will back to the official version. Again, you may need to restart your R session.

lavaan source code

All development of lavaan happens on GitHub. If you want to download (or clone) the lavaan source code, or if you simply want to browse through the source code, go to the lavaan github page:

https://github.com/yrosseel/lavaan

If you click on the commits tab, you can see all the recent changes that have been committed.

Development notes

Here, I will publish some notes that document important changes from version to version (or perhaps possible changes that are not implemented yet). I welcome any comments or suggestions on these notes.

• handling linear equality constraints in lavaan 0.5-18 (compared to 0.5-17 and lower)

Future plans

Too many. Eventually, we will get (most) of the features of LISREL/Mplus/gllamm/… in lavaan. But it takes time. Please do not ask me when a feature will be ready. I have no idea.

Features that are planned for future updates within the 0.6 series:

• exploratory SEM (ESEM) and exploratory factor analysis (EFA) (available in 0.6-8)
• two-level SEM with random slopes
• fiml for two-level SEM (to handle missing values) (available in 0.6-9)
• accelerated EM
• structural after measurement (SAM) estimation strategies (available in 0.6-9)

Other plans for future releases:

• better support for models with NO latent variables
• non-iterative estimators, James-Stein type shrinkage estimators
• improved small sample inference
• multilevel SEM with categorical data (using adaptive quadrature)
• support for discrete latent variables (aka mixture modeling, latent class analysis)