[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Stata (gllamm) benchmarks for different platforms?
Ronán Conroy wrote about -gllamm-:
> It's a major piece of intellectual work, and is
> really well supported with a website (www.gllamm.org) documentation etc.
> Sophia, Andrew and Anders are doing a remarkable job and perhaps more need
> support than to have their software made official.
> -gllamm- is the main Stata competitor to SAS Procs Mixed and GenMod. These
> are tools that are essential in many fields, and having a fast competitor is
> very important. Not having one loses sales to SAS... not what StataCorp
> wants, I'd hope.
I couldn't agree more and to quote myself: "-gllamm- is a fantastic piece of software". But I want to add, that it is really much more than a software:
-gllamm- is the practical implementation of a complete theoretical framework of "Generalized Linear Latent and Mixed Models" (GLLAMM) with capabilities of analyzing random effects models, marginal models, multilevel models, confirmatory factor models and structural equation
models (as well as any combination of them) and it comes with a link function for continuos, binary as well as ordinal data.
All this within a single Stata program!
The more I learn about GLLAMM the more impressed I become of it's generalizability, flexibility and pedagogical qualities. Sophia Rabe-Hesketh, Andrew Pickles and Anders Skrondal's work has meant a lot to me, by helping me to understand all these models within a single theoretical framework. And I would like to express a special thanks to Andrew Pickles for his excellent training course in Cologne this spring. If you have the opportunity to attend one of his lectures, don't hesitate to do so.
However, -gllamm- is very computationally demanding. It means that the practical potential (unfortunately) is far less than the theoretical potential. Improving execution speed is essential in increasing the practical potential. And by that a more common use of this fantastic piece of software.
I think, the authors of -
gllamm- chose Stata as their development platform for very good reasons. I know that StataCorp has been working with the -gllamm- team to improve performance and that part of their work was included in the executable some time ago with a boost in -gllamm- performance. This work, I'm sure, will continue. And I'm sure also, the fine tuning of Stata code for the Highest Performing Computer platforms.
BTW, -gllamm- is constantly evolving. I noticed they released a new version (2.3.5) on April 16.
* For searches and help try: