Know any good online resources about Structural Equation Modelling (with AMOS)?

Did you Google AMOS SEM? There are a variety of sites and lectures that various professors have placed online. Here’s one. There are many others and you could look for one that hits the level you want.

The book I have in my office is “Principles and Practice of Structural Equation Modeling,” by Rex B. Kline. I haven’t read it myself, so I can’t say how good or how easy it is. If you are used to regression modeling, then SEM isn’t all that different. It uses Maximum Likelihood algorithms instead of the Least Squares method. It’s purpose is to infer the effect of factors that you can’t measure directly (latent variables).

It’s also good at missing variables inference. It has a variety of methods for that.

It’s kind of cool because it has a visual way of diagramming your models, which is really useful for visual learners.

Of course, I’m kind of putting the cart before the horse. I’d have to know what your research really is all about, just to make sure that SEM is what you need. There are different varieties of SEM—I think both SAS and Stata have modules that let you do SEM.

Anyway, if you’re working on your PhD, and you need to use this method, then I wouldn’t try to fake my way through. I’d read the book and learn it for real.

And I am not a stats nerd!!!!

Thank you so much for your response, that’s really helpful!

Yep, I did have a Google around, but I just thought I’d have a check to see if anyone could recommend any particularly good, reliable resources. Someone recommended to me David Kenny’s website, which seems great, I’m just asking to see if anyone has anything else they could recommend too – I find it helpful to try and learn from a variety of sources, just in case one thing explains something in a way I don’t understand.

Sure, I totally understand the importance of understanding it properly rather than just blagging it, but it’s just helpful to have a look at an “idiot’s guide” before diving into the rather dense (but very useful) book that I’ve been loaned!

Based on my current understanding, I think SEM seems like an appropriate way of analysing this data. I’ve already constructed a diagram of my theoretical model, mapping out the various possible moderation and mediation relationships, and would like a statistical method which allows me to look at the different effects of a large number of variables all in one go, rather than conducting a series of separate analyses. I have a large sample and so think so think the SEM should be adequately powered.

Do you think you’ll run into a bunch of multi-collinearity issues? Would factor analysis be useful? Not knowing what you’re dealing with, I can only guess. From what I understand, making huge complex models is a rookie mistake. I think it can become very hard to interpret the results—perhaps you lose a lot of significance, even if you do have a large sample. Simplify, simplify, simplify.

You need to think pretty carefully about what you are doing (well, duh). Or maybe not so duh. In my experience, grad students will often bite off a lot more than they can chew. The other think I’ve seen often is that a committee chair will suggest a technique, not really understanding it, and the technique turns out to be inappropriate. Yet, the grad student feels obligated to use the technique because their chair suggested it.

Trust your own thinking. You’re the one who is doing this. Don’t assume those who are more experienced know what they’re talking about. It can be really helpful to have someone with more experience look at your model even before you try it out.