# Applying psychonetrics to Compare Psychometric Factor and Network Models

**r – psychonetrics**, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)

Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

This blog post was written by Kees-Jan Kan.

The first publications using R package *psychonetrics *are popping up!

In a recent publication and its associated tutorial website, we (Kan, de Jonge, van der Maas, Levine & Epskamp, 2020) illustrate how *psychonetrics* can be used for both Gaussian graphical modeling (GGM) and latent variable analysis. This shows, first of all, the enormous flexibility of *psychonetrics. *This flexibility comes with another advantage: researchers can now compare the fit statistics of psychometric network models and latent variables models without worrying if differences in results might stem from the use of different software packages.

A comparison between the network and the latent variable models is especially valuable (‘fair’) when both types of models are truly confirmatory. Our paper therefore included such comparison. For those who are interested in the subject, intelligence, we have found the following results: Network models of intelligence replicated over (WAIS standardization) samples, like well-established factor models do. Furthermore, the confirmatory networks provided considerably better fits to the data than those factor models considered. Together, these results thus provide (further) support for a network approach towards general intelligence.^{*} Because our paper is accompanied by a short *psychonetrics* tutorial of how to conduct such series of analysis, we hope you can use part of the code for your own model comparisons, or find other inspiration in it.

Implicitly, our paper illustrates that *psychonetrics* can be used for all kind of analyses that parallel those we are accustomed to in latent variable analysis and structural equation modeling. Think of introducing (equality or inequality) constraints to your network models. Equality constraints make it possible to fix the edges to specific, user provided values, or to test if certain edges are equal to each other, for example. This would come in handy in multi-group analysis, e.g. when you would want to test if (part of the) network’s structure can be considered equal across the sexes, age groups or other subsamples. With inequality constraints, you can test, for example, if all (or certain) edges in a given model can be considered positive.

In that sense, our paper on intelligence can be considered as applying some of the functionalities of *psychonetrics* that were overviewed in Sacha’s psychometrika publication (Epskamp, 2020).

We are looking forward to more papers!

* In our previous work, we used R package *OpenMx*, which is also flexible of course, but in our view less user-friendly, partly because this package has not specifically *designed* to conduct GGM.

**leave a comment**for the author, please follow the link and comment on their blog:

**r – psychonetrics**.

R-bloggers.com offers

**daily e-mail updates**about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.

Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.