What’s New in SPSS/ PASW/ IBM ? & Do I care?

Went to the “See it in SPSS” event at the Westin Bonaventure this week and, as always, it was well done. The two most interesting parts, based on a non-random sample of one, i.e., me, were:

• What’s new in SPSS (besides the fact that it is now called PASW and owned by IBM)?
• And the PAWS Modeler.

The new PASW 18 runs on Snow Leopard on the Mac (big plus for Mac people). Doesn’t run on the earlier Mac OS’s (big minus for Mac people).

There is a new boot-strapping option for some statistics, which is cool, but costs extra, (by which I mean a lot extra, according to the SPSS website)

One interesting addition was better integration with R. This will get me thrown out of the “in-crowd” with statisticians, I know, but I have never had much need for R. I have everything I can handle with SPSS, SAS and the occasional Stata, less occasional JMP and rarer S-plus. However, if you are not me (which I am fairly certain is the case) maybe you are interested in it, so there you R.

There is a Direct Marketing module, which is also cool and has applications far broader than marketing, but also costs extra. According to the SPSS website they had this in 17 but I have never seen it before.

When I saw the multiple imputation option I thought that was part of PASW-18 but when I installed the faculty pack, there it was, as part of the Missing Values module.

Speaking of the Faculty Pack – that is not new but it’s an awesome deal ($250 for  a one year license for 12 modules – 13, if you have Windows). It’s an awesome deal compared to buying each module and if you teach using SPSS you should definitely get it. I am going to spend some of my weekend and the next two weeks when I am laid up in the hospital playing with it.

Predictive Modeler I kept thinking looked a whole lot like Clementine and could not quite figure out the difference. Then I read that PASW modeler was formerly Clementine. It has all of the pluses and minuses of data mining software in general.

One question I asked the presenter which intrigued me was how long it took to install. He said an hour and a half. I have NEVER been able to install SAS Enterprise Miner on any computer.  Part of this is no doubt me. I have an extremely long list of projects that really need to get done, like updating our statistical consulting website at the university, writing new courses, re-writing a program that creates static web pages to be dynamic, possibly using SAS IntrNet (which requires learning more about it), doing programming for consulting clients, re-designing client websites to have RSS feeds from their employee blogs, etc.  It may only take 8 hours to install Enterprise Miner but I can’t see when in the foreseeable future I will ever have eight hours to spend installing something I really have no current need to use and am just interested in.

This is an interesting barrier for technology products, I think. There are things like R and the open source qualitative analysis program that all of us at lunch remembered trying and no one could remember the name. These are free. I know I won’t have time to learn either of those in the foreseeable future unless someone invents a machine to give you 48-hour days. (THAT I would buy! ) There is S-plus which looks interesting and I can pretty much bet I will never have time to learn. I have had the new SPSS modules for a few months and am finally getting time to look at those. I have been meaning to learn JMP for a year and have opened it and spent an hour or so looking at it, as well as attending an event sponsored by the JMP folks. I could go on … here is my point … all of this stuff is free to me because the university pays for the licenses, or it is open source. If you can’t get people to use your stuff even if it is free … well, that is a pretty tough market to penetrate.

What’s the secret? If I ran the world, I would start more of these events with “Here is how you get it installed and working on your computer, connected to your server, etc.” Don’t get me wrong, I LOVED the stuff on Bayesian statistics that Bob Rodriguez talked about at WUSS and I really liked the person from SPSS whose name I so ungratefully forgot, who discussed using equal N’s in groups (e.g., died/ not died) in creating an equation, for the training set, but proportional groups in the validation set because it is one of those simple, brain-dead obvious things that is too often forgotten (including by me). There needs to be a balance. I am busy and so is everyone else I know, except for my 18 month old granddaughter and my 76-year-old mother, and neither of them needs statistical software.

Unfortunately, we don’t have a license for PASW Modeler, so I am not sure how to test the claim it can be set up in 90 minutes, including installation of PASW Statistics. I have to say that I am intrigued.

My point is, even if your software is amazingly amazing, if it takes days to get it configured, no matter how great you are, people just won’t bother.

It reminds me of years ago when a relatively young woman offered to work for a company where  I was a partner. She presented herself to the only male partner and told him that she would be “willing” to work for us and named what I considered an astronomical sum, since it was more than any of the three partners made. (Whatever book she had been reading about job-hunting, I hope she got a refund.)

I asked my partner,

“You interviewed her. What did you think? Was she worth that kind of money.”

He said,

“Let me put it like this. I wouldn’t pay her that if she #$@%ed me with my morning coffee.”

My point, and I do have one, is that no matter how awesome you might be, you can price yourself out of the market, whether it be in salary or just the pain in the ass of getting the thing to work. From what I saw of PASW 18 this week, the pain in the ass cost is really low.