On 4 Sep, 16:23, William Stein <[email protected]> wrote:
> I'm just curious:
>
>    (1) what is "structured equation modeling"?

Basically, it is a way to display covariance. You build a model based
on covariance which may help to explain cause and effect. Of course,
you have to design the model very carefully and try a lot of different
models before you find a suitable model for your data, and this can be
rather cumbersome if you have a lot of different variables that you
want to combine and see the results of.

For more information, you could check out
http://socserv.mcmaster.ca/jfox/Misc/sem/SEM-paper.pdf
This paper deals with using R for SEM. It also explains a few of the
basic concepts.

http://en.wikipedia.org/wiki/Structural_equation_modeling
The wikipedia page which deals more with semantics.

http://apm.sagepub.com/cgi/pdf_extract/30/3/259
One of the many books available on the subject.

>    (2) Could you discuss more how/why R doesn't meet your needs?  I'm
> very curious what the Sage project might be doing right for you that R
> isn't doing.

When I was doing SEM using R for my master thesis, I found it rather
difficult to work with. Just finding the package for R and installing
it took some time, then trying to work out the somewhat lacking
documentation and figuring out how to use the different parts of the
package was far from easy. When I finally figured out how to work with
the thing, plotting the graphs left a lot to be desired. I ended up
plotting the graphs in the very basic form that R provides until I got
the model I wanted to use and then I had to re-draw the chart manually
using Inkscape to get it in a form that I thought was presentable.

Basically, the SEM package for R isn't complete. And it also has a few
bugs in it, which means that negative values for covariance produces
very awkward results. These bugs are acknowledged by the author of the
package, but as far as I know there hasn't been any development for
quite a while. The homepage of the author (jfox, linked from
http://cran.r-project.org/web/packages/sem/index.html) isn't even
available any longer. Even with the issues of bugs and rough
documentation and difficult interface and so forth, the basic
functionality of the SEM package for R is working well enough to
present some data. But the options for producing good graphs from the
data is just not acceptable in my opinion.

My master thesis was dealing with the possible roots of attitudes
toward bears in Sweden. You can find my thesis at
http://www.dan-erik.com/attitudestowardsbears.pdf . When working on
that project I found SEM to be a very nice way of visualizing the vast
dataset that I had. Now I am working with research on fish with a more
biological approach to issues, but I still think that SEM could be a
good way of visualizing some of that data. Working with the R
environment and especially the SEM package for R is not something I am
looking forward to after using Wolfram Mathematica and sagemath for a
while. I find these tools to be a lot more intuitive when coding as
well as providing better options for graphs and visualization. So I
was hoping to find a way to work with SEM through those tools rather
than R.
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/sage-support
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---

Reply via email to