Oh, it's SEM contemporary!
My learning path into R-SEM-Goodness
OpenMX RSS Feed
0Well, this was tricky. I wanted to get the news from the OpenMX project. On their homepage you cannot subscribe to a RSS feed, but clicking on more in the News section, you come to the blog of the OpenMX project. Alright! There we are: http://openmx.psyc.virginia.edu/rss.xml is the Feed address. Let’s put this into our reader:
Now we will see new releases from OpenMX in our RSS reader without the need to visit the site every day!
Next steps
0OK, when talking to Mark I noticed the ggplot2 book on his desktop – ggplot2 is an awesome graphic package for R – just when I thought lattice was cool. Distraction, distraction… but I decided on the spot to include ggplot2 in this semesters learning project (I also signed up for the ggplot2 mailing list, of course). I will post my ggplot2 ramblings on my general R blog, so hop over to see World of R-Craft. But I will make any postings I do on that other blog visible by setting up an RSS feed in the sidebar of my blog.
First steps – setting up mailing lists
0Although a lot of people nowadays don’t know what they are, mailing lists are a great resource for getting help in your learning project. But before everyone posts some already answered questions on a mailing list, here are some tipps:
First, I searched for relevant mailing lists. I found the
Read the Posting Guide of the mailing list before posting! For example, have a look at the posting guide of the r-help list. To paraphrase the most important rule relevant for every mailinglist on the planet:
Do your homework before posting: If it is clear that you have done basic background research, you are far more likely to get an informative response. See also Further Resources further down this page.
This means: before you post to a mailing list, you should have spent considerable time thinking about and trying to solve the problem by yourself. Considerable time means at least a few hours, if not days. And by ALL means, do a search of the mailing list ITSELF before posting. This can be done by accessing the web frontend of the mailing list, for the r-help list head over to R mailing lists archive search.
It is also a good idea to search for your relevant topics in the mailing list archive to get a feeling for the relevance of the list to your specific interests. I searched for “sem package” and got 399 hits. OpenMX only 9! Ouch! Is there a OpenMX specific mailing list? Mmmh, let’s look at their website… no, but there is a forum and also a wiki. OK, so it becomes clear, that the r-help mailing list is NOT the right place to ask OpenMX questions.
Mailing list are notoriously inbox flooders (= there are a lot of emails coming in and fill up your email inbox). My solution to this has been that I have set up a new email account and set up some mail filters. Here is my first few hours of incoming mailing lists email with some filtering in Gmail:
Next steps will be to collect a lot of good RSS feeds on SEM and OpenMX.
Start into SEM with R
2First steps should be easy. I am going to look at R packages for doing SEM analysis.
In the CRAN Task View: Social Science it list following SEM packages:
Structural-equation models: The sem package fits general (i.e., latent-variable) SEMs by FIML, and structural equations in observed-variable models by 2SLS. Categorical variables in SEMs can be accommodated via the polycor package. The systemfit package implements a wider variety of estimators for observed-variables models, including nonlinear simultaneous-equations models. See also the pls package, for partial least-squares estimation, and the gR task view for graphical models.
So obviously John Fox‘ SEM package is the first stop SEM package for R.
But there is also OpenMX – this looks very promising:
OpenMx is free and open source software for use with R that allows estimation of a wide variety of advanced multivariate statistical models. OpenMx consists of a library of functions and optimizers that allow you to quickly and flexibly define an SEM model and estimate parameters given observed data.
OpenMx runs on Mac OS X, Windows XP, Windows Vista, and several varieties of Linux. This means the same scripts you write in Windows will run in Mac OS X or Linux.
OpenMx can be used by those who think in terms of path models or by those who prefer to specify models in terms of matrix algebra. OpenMx is extremely powerful, taking full advantage of the R programming environment. This means that complicated models and data sets can be specified and modified using the R language. In order to give a very brief idea of what OpenMx looks like, here are two small demo examples: one from a path modeler’s perspective and one from a matrix algebra perspective.
Great! So what I will do is that I am going to dive into these packages first and do some basic SEM stuff. All I need now is some good data to test my stuff. Often R packages have some sample data sets, so I need to dive into the documentation (SEM, OpenMX). There is also a course by John Fox online, so I will have a look at that.
I also have the workshop notes from John McArdle, but clearly I have to cover the basics of SEM with R first before I get into the more advanced latent growth models.