Dynamic Structural Equation Models (Asparouov, Hamaker and Muthen)

Hi all. This doesn’t matter to me at all, but it might interest some of you: https://www.tandfonline.com/doi/full/10.1080/10705511.2017.1406803

Dynamic Structural Equation Models (Asparouov, Hamaker and Muthen)
This article presents dynamic structural equation modeling (DSEM), which can be used to study the evolution of observed and latent variables as well as the structural equation models over time. DSEM is suitable for analyzing intensive longitudinal data where observations from multiple individuals are collected at many points in time. The modeling framework encompasses previously published DSEM models and is a comprehensive attempt to combine time-series modeling with structural equation modeling. DSEM is estimated with Bayesian methods using the Markov chain Monte Carlo Gibbs sampler and the Metropolis–Hastings sampler. We provide a detailed description of the estimation algorithm as implemented in the Mplus software package. DSEM can be used for longitudinal analysis of any duration and with any number of observations across time. Simulation studies are used to illustrate the framework and study the performance of the estimation method. Methods for evaluating model fit are also discussed.

I’m guessing it would be much simpler, as well as more flexible and reliable, to fit this model in Stan, which could benefit whoever is out there doing this sort of thing.

[edit: escaped quote]


This one will be easy to ignore, it’s the most expensive paywalled paper I’ve ever seen!

1 Like

Someone sent me the article itself, if you’re interested. See attached. I didn’t read it myself.

edit: [copyright violating link removed]

I removed the link as it’s a copyright violation. This is why I hate paywalled papers—you can’t legally give me a copy.

OK. If anyone is interested in programming this model in Stan, just send me an email and I can send you the article.

Please don’t offer to violate international copyright law on our mailing lists.

Fair enough. The first author of the paper is Tihomir Asparouhov, tihomir@statmodel.com. You can email that person and maybe he or she can send you a preprint.

[edit: I misunderstood—Andrew was talking physical preprint. My bad.]

You seem to be missing the point here, @andrewgelman. It is not legal for people to send other people copies of copyrighted papers that are not licensed for free distribution.

I do not want to encourage illegal activity on this blog. I’m taking down that last link so you don’t get someone else in trouble.

Oops! Went to edit it and realized you said “preprint”. I didn’t even realize there still were preprints. The NLP and ML journals went online ages ago.

I am not suggesting that anyone send other people copies of copyrighted papers. I’m saying they can send an email to the authors asking for a preprint. Not the copyrighted version. Whatever.

@andrewgelman, thanks for clearing that up. @Bob_Carpenter, thanks for activitly moderating the forums.

For the moderators: please continue to discourage illegal activities on the forums.

There’s two different senses of “preprint” in play here.

The original sense is the old-fashioned one where a journal would send you a stack of copies of the typeset paper for publication before it was published. Those were subject to copyright and you couldn’t legally just photocopy them and distribute them to everyone you knew or scan them and put them on the internet (which was a thing before the web was a thing for those who didn’t realize).

The modern sense is that of putting a paper on arXiv before submitting it. Or of posting the submitted version to satisfy NIH requirements when journals won’t let you distribute the copyright version.

Either way, those are legal to send to people.

Because of the variability among publishers, what Bob said is only partially correct.

A few years ago this was the picture:

87% of journals allowed immediate self-archiving of some version of article (pre-print, post-print, published) – therefore 13% of journals did not allow self-archiving of any version of the article (pre-refereed included);
60% allowed immediate self-archiving of post-refereed version (aka post-print);
16% allowed immediate self-archiving of published PDF.

A quick thing to do to know what we can do with our published articles is to look here:

For the journal in question (Structural Equation Modeling: A Multidisciplinary Journal), we find that the authors can archive both pre- and post-print versions.

The freedom to share the post-print version with no embargo is something that we should always demand. If the copyright document indicates that we do not have the right to self-archive the post-print, we can submit an amendment (they may say no, but we should always try):


That looks useful. For the Structural Equation Modeling Journal, it says the author cannot archive publishers version/PDF (I thought that was the link I was taking down). There are also a set of side conditions listed, such as that it must include a link to the published version. They do seem to allow posting on a non-profit server, so I suppose this would qualify if the link were to a post-print.

By archive, do you mean they can post them on the web for others to download? Can they do that on sites other than their own university or home pages?

That’s a good idea. Most people don’t realize that there’s room for negotiation in these situations. Probably not with every journal.

agree, but I would suspect that you are legally responsible only if the file is stored on your server. If that was a link to a file stored somewhere else, the people who archived it are responsible. In any case, I understand that as a moderator you do not want ambiguities and prefer to adopt a conservative approach.

My understanding based on conversations that I had a while ago with copyright people at my institution is that, unless otherwise specified, archiving refers to public sharing on any non-commercial website. Rules are usually (probably almost always) different if the article is used for commercial or marketing purposes.

Uploaded files are stored on our servers (technically, Discourse’s servers—we don’t do our own hosting of this forum).

I’m totally OK with content that’s legally licensed for public redistribution.

I’ll remain conservative in the sense of expecting the poster to indicate the license allowing the content to be legally distributed.

That site of journal policies said they required non-profit status, not just non-commercial. Usually, licenses don’t distinguish between commercial/non-commercial (lots of people think GPL prevents commercial use, but they’re wrong—it’s about source distribution). Often academics try to write qualifiers on their distributions like “for academic use only” or “for non-commercial use only”, but it’s rarely clear what they mean by that (e.g., what if somone at Google uses it to write a NIPS paper? what if someone at Columbia uses it to write a book that they then sell?).

I think we’re OK with the non-profit requirement being under the NumFOCUS umbrella.

Those are also different than being a non-profit. It’s very confusing, but non-profits can sell things—Columbia university sells degrees, for example, and Columbia professors sell books (and journal articles!). And they engage in a ton of marketing. Nevertheless, Columbia’s a registered non-profit.

Pretty much everything can be construed as marketing, including these forums, Andrew’s blogs, the Stan web site, etc.

Here is the link to this and other papers using mplus on their website: http://www.statmodel.com/papers.shtml