After reading David Smith’s tweet on the price of Oracle R Enterprise (actually free, but it requires Oracle Data Mining at $23K/core as pointed out by Joshua Ulrich.) I went to Oracle’s site to see what was all about. Oracle has a very interesting concept of why we use R:

Statisticians and data analysts like R because they typically don’t know SQL and are not familiar with database tasks. R allows them to remain highly productive.

Pardon? It sounds like if we only knew SQL and database tasks we would not need statistical software. File for future reference.

Alternatively:

Database professionals like SQL because they typically don’t know R and are not familiar with a real programming language. SQL allows them to remain highly paid.

Disgruntled coder?

Lol-worthy!

Of course, all you have to do compute the variance in SQL is figure out how to do this:

SELECT POW( SUM( POW( data, 2) – AVERAGE(data) ), .5) / COUNT(data) AS variance

See how easy it was???

So much work. Try this:

http://docs.oracle.com/cd/B19306_01/server.102/b14200/functions212.htm

Wow, there is a variance function and one can even run a STATS_ONE_WAY_ANOVA!. And how come that Colmenares gets such a crummy salary?

Which is not to say that I’d rather use a database’s SQL variant of stats, just to be clear.

Fact is, SQL standard has been adding ever more stat functions with each version. Their implementation is not uniform across engines, however. For those who want to mix the two most tutorials (usually early on) draw the parallel between data.frame and relational table. Merge is just a kludgy join. And so on.

Then, there’s my beloved PL/R. If only it ran on DB2. I’ll need to look some more at this Oracle/R marriage. First announcements were ambiguous; is it PL/R for Oracle, or just another R driver for Oracle.

Nobody’s going to jump on the brash comparison between R and Excel?!

Can Oracle generate directly the kinds of graphs that R can produce or is that an extra cost ?

Depends on how you measure “directly” and “extra cost”. If you follow the original link back to the Oracle page, you’ll find that R support is “free”, but requires a non-vanilla version of Oracle. Part of the support includes the ability to run R in Oracle similarly to what PL/R does in Postgres; so the answer is “sort of”. That said, the emphasis is on using Oracle within R, for various benefits.