A flurry of press releases from SAS®.

Two caught my eye: SAS to offer R integration to support analytical innovation, and SAS and Teradata Announce Availability of Analytic Advantage and Optimization Services Packages.

SAS to offer R integration to support analytical innovation

SAS, the leader in business analytics, is expanding analytical options for its customers with a new interface to R open-source statistical software. SAS’ initial integration with R will be included in SAS/IML ® Studio 3.2 (formerly SAS Stat Studio) scheduled for summer 2009.

…

“It is no secret that SAS has been working on interfacing with R,” said Anne Milley, SAS’ Senior Director of Technology Product Marketing. “SAS and R are here to stay, and as organizations work to harness the full potential of their data, an expanded set of analytics options can only help.”

So if you want the R interface you need to license software (SAS/IML) that under-performs the software (R) that you currently use. I think that this is a brave step by SAS, but it looks a little misguided: the power of R is that it interfaces with SAS data and provides functionality not available in SAS and all of this for little cost (cost as effort, rather than just license dollars).

If R was being used as a marketing tool to try to sell SAS, I could imagine an antipathy or backlash developing in the R community.

SAS and Teradata Announce Availability of Analytic Advantage and Optimization Services Packages

Why is it that when we’re paying for SAS and Teradata we feel that we can expect to pay for their integration, but when one component (eg R, see above) is free, we expect the package should be free?

“Our partnership with SAS creates a highly innovative dream team focused on helping companies activate and optimize analytic model development and execution inside the database, where we see processing speeds increase as much as 45 times,” said Rob Berman, SAS Partnership Vice President for Teradata. “These two packages are just the beginning of an in-database planned portfolio of offers we’ll deliver to a market that is ready and waiting for it. In view of the economic pressures companies now face, these packages have come at exactly the right time.”

…

The Teradata and SAS partnership has continued to earn high marks from the analyst community. “Both SAS and Teradata are large, financially stable companies with loyal customer bases,” said Helena Schwenk, Senior IT Analyst at Ovum, in a new analysis. “More importantly, both companies’ software, applications and infrastructure are essential tools for helping companies beat the downturn – whether they are used to highlight cost reduction opportunities, improve supply chains, optimize inventory levels, help businesses understand risk or rapidly changing customer preferences and spending.”

I look forward to hearing more!

Technorati Tags: analytics, integration, r, SAS, teradata

Well, I couldn’t quite believe how simple the syntax is, it feels quite strange using <- all over the place, and not a proc in sight (R exposes functionality through functions).

Moving very slightly forwards from the intro code I wrote, with a bit of trial and error, the following:


library("foreign")
testsas <- read.ssd("c:/temp", "test", sascmd="c:/program files/sas institute/sas/v8/sas.exe")
attach(testsas)
summary(testsas)
fm<-lm(y~x,testsas)
summary(fm)
plot(x,y)
abline(coef(fm))


And it worked!

The "exciting" thing about this rather mundane code is that the read.ssd() function is reading a SAS® dataset into R! Reading a SAS dataset is no more difficult than loading the "foreign" library, and then passing the path, the dataset name (without the .sas7bdat extension), and the path to sas.exe.

The dataset is fairly dull - it has two variables x and y, where y is x plus x times a random amount between 0 and 1.

The attach() function makes the dataset available as a data frame, the summary() function gives basic summary stats, the lm() function is equivalent to proc reg, and stores the regression coefficients to a new variable structure fm.

Another call to fm() gives me linear regression coefficients (the m and c, from y = mx + c) and goodness of fit statistics.

Finally the calls to plot() and abline() plot a chart of observations, overlayed with a line of best fit.

Not very interesting but: (a) reading SAS data into R is quick and easy; (b) writing R code is quick and easy; (c) the results from the R code were verified by those produced in SAS.

Point (c) isn't much of a surprise - but the painlessness of both (a) and (b) is!

I look forward with great interest to hear what comes out of the SAS Global Forum in a few days time!

Technorati Tags: data, r, SAS

It was interesting to read about SAS’s use of open source software:

Open source software provides many options and benefits. We see many (SAS included) embracing open source for different things…we plan to do more with open source in the future.

But nothing more definite yet, we’ve got to wait for SAS® Global Forum:

The first step of SAS integrating with R will be shown at SAS Global Forum coming up in DC later this month. Other announcements for new offerings are also planned at this event.

We will wait with baited breath, and continue on the path towards R, happy to have Anne Milley’s full backing:

I have great respect for people who invest in learning (or even getting exposure to) more than one language and who appreciate the strengths of different languages for certain tasks and applications.

I only wish I could go!

I wonder if there will be any meaningful talk there of the longer term plans for SAS Institute, a topic avoided in the interview.

Technorati Tags: r, SAS

Yet with our annual subscription licensing model, SAS cannot rest on its laurels. Each year, customers vote with their checkbooks: if SAS provided them with business benefits, results and a positive ROI, they renew; if not, they can walk away. Happily for SAS, the overwhelming majority of customers keep coming back. But the licensing model keeps SAS on its toes, customer-focused, and always listening and innovating based on customer feedback.

Is an interesting take on it.

The debate, guessing, second-guessing, spinning and counter-spinning continues. Doom-mongers always come out on top in situations like this: if they’re right, they’re heroes, but if they’re wrong their errors are forgotten and they can go on to make further ominous predictions.

If only SAS® could swallow up R, this whole conversation could quietly go away. As it is, with R in the public domain, that’s not an option. It will be interesting to see how SAS responds, and I think there are a couple of likely responses:

Rely on the quality proposition – sure, it’s more expensive, …, but its better (Stella Artois)

Embrace open-source for the greater good of the analytics community, realising that if they can’t grow their market share, they can grow the market (IBM model)

But it all depends on what the longer term vision is: what will SAS look like in ten years time.

Will Dr Goodnight still be at the helm? If he’s not, will somebody have bought SAS – I can’t imagine SAS being bought by Microsoft – but who else has deep enough pockets?

I suppose these short-term issues could have a dramatic influence on the purchase price of SAS Institute: if licensing revenues fall and the value of the IP is eroded by open-source offerings, then what is left other than an estate agency? How deep would those pockets need to be?

If SAS weren’t sold, could the SAS Institute become a self-owned, or employee-owned, foundation?

I’m curious to know more about the long-term perspective: how does SAS see itself in a decade, it would be great if part two of Anne Milley’s interview gave us a taste for this.

Technorati Tags: future, r, SAS

And then I read that R is the most popular statistical package at Google. Clearly not so bijou, after all.

So now the tricky question: if we are going to embark on an exploratory data-mining exercise, do we go down the SAS® route and license Enterprise Miner, or do we jump on the band wagon and try a proof of concept using R?

I hadn’t realised, for example, that R had packages that allow one to interface with SAS datasets – this means we can exploit data we are familiar with no need for data conversion…

… but, I’m the SAS admin? Part of this feels very wrong. Perhaps I should rebrand myself as an Analytics Support Manager? Not as snappy, I’ll grant you.

Addendum: After reading the interview with Anne Milley, I laughed out loud when I saw Ajay Ohri’s earlier post (R for SAS users,) just what I need!

Technorati Tags: analytics, r, SAS

After posting about charting a route through a rolling biblioscape yesterday, my mind turned to the route I’m treading at work. Is SAS® the high-point? I hope so, but the recent interest in R got me thinking.

R is a closed book to me, I’d never come across it before. Finding a new area of complete ignorance is always exciting. I decided to start investigating.

My first port of call was the R Project site. Here I was able to quickly download and install R on my desktop machine and find good introductory documentation which included a sample session that I could walk through.

I was pleasantly surprised by how easy it was: the installation was trivial and the sample code worked easily. I was interested by how terse R syntax seems to be, and how quickly one could create output. And all this with anonymity: no registration, no download fees.

So the question is: is R on the plateau around Mount SAS, or is SAS merely base camp for the ascent of R?

I haven’t a clue yet, but I’ll let you know my thoughts as they occur.

Why the interest in R? It is part idle curiosity, and part a desire to hedge my bets. I wouldn’t have backed Linux against Windows, and yet look at the success it has enjoyed: around 14% of the server market. Having seen this rise from zero to hero, anything is possible. Especially in the current economic climate. If I was starting out, I’d be very tempted to take a punt on R – there appear to be immediate savings in both time and money when compared to a front-loaded license fee model. If students are using it then there is potentially a pool of able users that could be hired economically and who would grow with the software as it continues to mature.

One thing that would give me pause is the cost of ownership. The technical support may not be to the standard offered by SAS, but most of my SAS problems are actually solved by the SAS community on forums and listservs, and this is exactly where open source projects draw their strength from.

The other question, and one I haven’t an answer to, is whether R can help me with the business problems I use SAS to tackle. The worrying thing for SAS*, is that I may be able to answer it quite soon: now that I’ve got R installed on my desktop I can play around with it whilst I’m waiting for remotely submitted jobs to execute on our server. I can do this for free in spare time.

* I’m guessing – I’ve got no connection or insight to SAS other than as a customer/user.

Technorati Tags: analytics, future, r, SAS

I read on sasinct.com about possible mergers or acquisitions of SAS by SAP in a few years time after they (SAP) have bought Teradata…

… and then at the other end of the spectrum SAS users on SAS-L are thinking about integrating SAS and R – see recent coverage on the SAS BI-ogsource. I think Anne Milley has opened a can of worms, and it will be interesting to look back in six months, or a year or two, and see what, if any, impact this has had…

…, whilst in another part of the forest Chris Hemedinger is trying to ram EG down our thoats with the support of various other SAS mouthpieces.

If we wanted EG, we’d be using it already, we can see that it is cheaper per seat than Base SAS.

I suspect this last is, in part, a response to the “criticism” from Gartner Group; their recent survey of the BI market placed SAS firmly in the “magic quadrant”, but had this to say about EG:

… anything beyond simple ad hoc reporting requires the power user-oriented Enterprise Guide, with many of the data manipulation and advanced analysis tasks requiring SAS programming language knowledge. The need for specialized skills represents a significant barrier to adoption in new SAS accounts.

Technorati Tags: future, r, SAS, sas-l, sasinct