Jorg,

Mplus can implement two levels easily, and up to three or four depending on the details of the analysis. Simply accounting for clustered sampling is quite easy. If you have variables of interest on multiple levels (especially a cluster level) it gets trickier. It's also not cheap.

If you're an R user, you might check out xxM at http://xxm.times.uh.edu/ , though I haven't yet used it myself.

Pat

Jorg,

Mplus can implement two levels easily, and up to three or four depending on the details of the analysis. Simply accounting for clustered sampling is quite easy. If you have variables of interest on multiple levels (especially a cluster level) it gets trickier. It's also not cheap.

If you're an R user, you might check out xxM at http://xxm.times.uh.edu/ , though I haven't yet used it myself.

Pat

Thanks, Patrick. This is quite helpful. As I am trying to get my head around R, I will follow up on the link you sent.

Great. It's not in CRAN at this time. Geoff Hubona pointed me to it.

Jörg, You may use the Mplus, Amos, and Lisrel software, but its related to your goals and in what way you want to used the SEM. Best wishes.

Jörg, I am going to concur that Mplus is a good tool for this sort of problem. Each major release of Mplus yields significant features. As long as you are currently licensed there is excellent support, too. There are excellent papers and other resources on their web site www.statmodel.com.

http://www.statmodel.com

Jorg, you might try Stata, which has improved its capabilities enormously in this area. The manual contains a very good introduction.

http://www.stata.com/manuals13/sem.pdf

Jörg, I work with STATA when conducting hierarchical modelling. STATA has also sufficient capabilities with SEM and can manage both ways of modelling jointly.

We had the same problem and our team used STATA 13 which turned super easy...the output looks like HLM and the Programming is really easy like AMOS. Its not that expensive , too!

Hi Jorg, I believe Gretl does it. Gretl is a free access and very powerful programme. you can download the programme at gretl.sourceforge.net/.

It is very straithforward to use and there is a lot of documentation on line. Good luck,

Michela

MPlus indeed...

As several others researchers have pointed out, go for MPLUS. If you are going to invest some funding, you will not go wrong with MPLUS.

The syntax language is easy to grasp, it handles several features, and it has a broad community of users; and recently implemented cross classified and multi-membership models,which are quite probable in any multilevel escenario.

If you want try other software, and you don't mind a bit of coding, R is good option. Though, as Damico says: it doesn't pay over night. R has many advantages, but these are in the long run.

MLwiN is quite versatile aswell, and has several tutorials at Bristol uni.

Check: http://www.bristol.ac.uk/cmm/learning/online-course/index.html

Good Luck!

R package lavaan might prove useful for analyzing Multilevel SEM.

There are also several Multilevel Modeling tutorials using SAS, STATA, SPSS and MPlus. For details, you can refer to this link:

http://ssc.utexas.edu/training/software-tutorials

I recommend you Amos and Lisrel software.

If you have budget constraints, you might also consider the new xxM http://xxm.times.uh.edu/ - it's a free R package.

xxM is a software package for multilevel structural equation modeling with complex dependent data structures. xxM implements a modeling framework called n-Level Structural Equation Modeling (NL-SEM) and can estimate models with any number of levels. Observed and latent variables are allowed at all levels.

I found EQS works best for me.

I would also recommend using Mplus. It is very flexible and has nice additional options (e. g. handling of missing data).

Mplus and you'll never go back. This might be useful http://www.ats.ucla.edu/stat/mplus/seminars/mlmMplus_JH/

Multilevel CFA or SEM not available in lavaan version 0.5-15 (15 November 2013). But to expect in lavaan 0.6 or 0.7. This a great package, outputs are similar to Mplus and EQS.

http://cran.r-project.org/web/packages/lavaan/index.html

Good Tutorial in:

http://lavaan.ugent.be/

If you have enough money and trust that the program does what it is supposed to do --the code is secret and the manual does not tell you or your statistics advisor exactly what it does-- take MPlus. If you are poor and you know how to work with R it's the package lavaan. I don't think the specification of the lavaan function is more complex that MPlus.

MPlus has the disadvantage that it's stand-alone and its language is specific for that program, lavaan is part of the statistical package R that has the advantage that its language transfers to other R functions. I'm not sure lavaan can do multilevel though. Check it out first.

I use both lavaan and Mplus. If money is not an issue then go with Mplus, the software is purely syntax based. If money is an issue then lavaan, but you may need to install a couple of functions

MPLUS can handle multilevel SEM. Is totally recommendable: versatile, easy to learn, eventhough is syntax based is friendly enough. The fullest version (combination add on) is at 350 USD for students, and 895 USD for universities. If you have never use it, "Data Analysis with Mplus" from Cristian Geiser, is a good book to start.

I agree with Diego. Also, the Geiser book only deals with continuous variables and not categorical plus there is no info on creating formative models, which is a limitation but its a good source for beginners. The same goes for some other sources.

I think MPlus or even Stata can do a good job. It depends on what you want to do with SEM

I really like MPlus, flexible, and easy to learn

The estimation method is affected by the normality analysis of your data. Software selection depends on your estimation method.