[R-SIG-Mac] Preparing mySQL queries for R and SNA

[elw at stderr.org]

> input in R.  I am hoping to do both the stats and SNA portions of my 
> analysis in R (using the sna package as well).  I endeavor to write a 
> script to handle everything.  My process flow is: mySQL >> R >> R(sna)
> >> output tables and graphs >> insert into docs.


This is indeed a sane workflow - pretty much what I use for my own social 
network analysis work.  I happen to use postgresql rather than mysql, but 
that is somewhat tangential to the problem and occasionally driven by the 
kind of data I tend to like to work with ;-)

Looked at the igraph package?

Side-tip -- the sooner one starts writing functions for big blobs of 
analysis rather than do-it-all scripts, the happier one seems to be.  A 
lot of my old code is not written as functions, and frankly I seem to be 
putting a lot of work into converting it TO functions prior to being able 
to reuse it.  Oops.


> Is .csv the best intermediate format for input into R and the sna 
> package, or will I have to write my own script to prepare the data in 
> special R(sna) matrix form?


As Byron suggests, pulling the data directly from MySQL into R is probably 
much simpler, and frankly is pretty straightforward.  You'll basically be 
writing selects that return vectors of data objects - pretty much what you 
likely already have in SQL, I would guess.

There is support for things like Harwell-Boeing sparse matrix format, but 
you can probably avoid the need for that.

The more you can do from within R, the more headaches you can save 
yourself.  R rules :-)


> I am analyzing online bulletin board projects, so each project (~1200) 
> will have a separate set of analyses done on it, then they will be 
> aggregated and summarized.


Nice scale.  Did I perhaps see one of your talks in Vancouver (at Sunbelt) 
a couple years ago?

Several people in my department do very similar work (CMC, SNA, 
computer-mediated discourse analysis, etc, ethnography, etc), and I would 
love to hear more about what you're doing.  Offlist, maybe?


--elijah wright
School of Library and Information Science
Indiana University, Bloomington