Structural equation models.

[noetsi]

I know SAS does this, Proc Calis. But I have heard doubts raised about how good that product is. Anyone know how good SAS is for structural equation models (L Learned it with Mplus, but there is no way we can use that here, they do not sell to the state).

 

[spunky]

To the best of my understanding, the only "bad" thing about PROC CALIS is that it doesn't quite keep up with the new developments and whatnot. But there's nothing bad per se about it if you're not doing anything super fancy.

 

[noetsi]

thanks spunky. Its a pretty safe bet I won't do anything fancy by your definition Or have kept up myself. I last studied it 5 years ago.

 

[spunky]

Quite frankly, given what you do and everything, I'm pretty sure that PROC CALIS will give you everything you need.

 

[noetsi]

thanks spunky.

 

[Dason]

I doubt it'll do what you need

 

[noetsi]

thanks for the vote of confidence dason

 

[Dason]

I honestly have no idea. I just like making your life harder. It's how I roll.

 

[spunky]

^^He's a known Cauchy-associate and shouldn't be trusted^^

Where are your (central) moments, Dason???

WHERE!?!?

 

[noetsi]

I spent a year in graduate school learning these models, but never used them (it does not help that I learned it in MPlus which of course is very different than SAS). For a long time I planned to stress multilevel models because most thought our areas were the key. But the results suggest they are not and I lack the software (given small group size) to run that anyhow.

So SEM and vector autoregressive models is what I plan to relearn. I wish there was a good time series regression method but after 6 years of effort in that regard none seem very useful.

 

[Dason]

Have you ever done var modeling?

 

[noetsi]

Have you ever done var modeling?

I have. I have never really understood these models, however notably how to interpret the results - the tools used are very different than standard regression. I am planning to try again - which was the point above.

I also need to learn cointegration to really use these as if cointegration exists VECM and not VAR models are required. I have worked with this, I need to do a lot more to truly understand them.