An SPSS refugee needs a stats package!
March 11, 2011 10:12 AM Subscribe
I've agreed to do a validation study for a client, so I'm looking for an inexpensive (under, say, $200) statistics package that will do descriptive stats, correlations, t-tests, and other simple analyses on a relatively small dataset (fewer than 500 cases). I went through graduate school and into my early career using SPSS and I was very comfortable with it. But I don't have access to SPSS now and it's just much too expensive for my little one-person company to purchase. Is there something akin to SPSS available? My statistical skills have gotten really rusty over the last several years so I'd like to avoid having to master a new style of interface (say, to use R or even Excel) while I'm also trying to get rid of the rust.
I'm afraid what you're looking for only comes in three versions: institutionally-priced statistics software, consumer-priced Excel, and R, which is free and starts out extremely simple.
The generic GUI that comes with R has very little in the way of interface, and you would need to learn very, very little code in order to do things like t-tests and correlations. Maybe "learn" is the wrong word. You would need to find a reference sheet and just copy the parts you needed. Even the one I linked to is probably overkill for your purposes.
In fact, you could write out all the meager code you'd ever need, just once, then save it in a plain-text file and run it over and over as necessary.
Numerous feature-rich GUIs are available for R. Most recently, people have been recommending R Studio.
Seriously, you're exaggerating the difficulty of what you want to accomplish.
posted by Nomyte at 10:31 AM on March 11, 2011 [1 favorite]
The generic GUI that comes with R has very little in the way of interface, and you would need to learn very, very little code in order to do things like t-tests and correlations. Maybe "learn" is the wrong word. You would need to find a reference sheet and just copy the parts you needed. Even the one I linked to is probably overkill for your purposes.
In fact, you could write out all the meager code you'd ever need, just once, then save it in a plain-text file and run it over and over as necessary.
Numerous feature-rich GUIs are available for R. Most recently, people have been recommending R Studio.
Seriously, you're exaggerating the difficulty of what you want to accomplish.
posted by Nomyte at 10:31 AM on March 11, 2011 [1 favorite]
Spend a couple of hours fiddling around in R before buying anything--you might be surprised at what you can accomplish. Take a look at the "Data Input" and "Basic Statistics" sections on Quick-R for sample code for everything you've described.
posted by esoterrica at 10:40 AM on March 11, 2011 [1 favorite]
posted by esoterrica at 10:40 AM on March 11, 2011 [1 favorite]
Numerous feature-rich GUIs are available for R.
These are glorified text editors, for the most part. Which is not to say they aren't useful, just that they don't hide too much away about the pain of writing working R code.
I would definitely recommend trying out R, because it is free and it is extensible. But I doubt you'll go far without a good intro book.
If you can spend the money, you might look into JMP, which is made by the SPSS folks and is somewhat cheaper.
posted by Blazecock Pileon at 12:39 PM on March 11, 2011
These are glorified text editors, for the most part. Which is not to say they aren't useful, just that they don't hide too much away about the pain of writing working R code.
I would definitely recommend trying out R, because it is free and it is extensible. But I doubt you'll go far without a good intro book.
If you can spend the money, you might look into JMP, which is made by the SPSS folks and is somewhat cheaper.
posted by Blazecock Pileon at 12:39 PM on March 11, 2011
I respectfully disagree. JMP is made by SAS. It's both expensive and, IME, takes a while to adjust to.
An IDE (a term I prefer to use instead of "glorified text editor") really helps when writing R code, and writing R code for simple tasks like correlations is not difficult. Really, it's a lot easier to write "cor(x,y)" in R than to click through several dialogs in SPSS. "The pain of writing working R code" is an overstatement when you're writing very simple code.
Finally, the Dalgaard book is a useful resource, but I find that it introduces too many R data structures too quickly and too unsystematically. A free PDF tutorial like Maindonald's Using R supplemented with a good reference card should more than suffice for a beginning user.
posted by Nomyte at 1:33 PM on March 11, 2011
An IDE (a term I prefer to use instead of "glorified text editor") really helps when writing R code, and writing R code for simple tasks like correlations is not difficult. Really, it's a lot easier to write "cor(x,y)" in R than to click through several dialogs in SPSS. "The pain of writing working R code" is an overstatement when you're writing very simple code.
Finally, the Dalgaard book is a useful resource, but I find that it introduces too many R data structures too quickly and too unsystematically. A free PDF tutorial like Maindonald's Using R supplemented with a good reference card should more than suffice for a beginning user.
posted by Nomyte at 1:33 PM on March 11, 2011
I've recently been learning R with R Commander, which has helped a lot as far as giving more of a point-and-click option (useful, since I'm trying to decide if it's worth teaching to undergrads). If you're looking to spend a little bit of money anyway, there are distance classes out there that can walk you through R for that kind of price point.
Are you near a college campus? If so, you're probably able to use SPSS on campus in the library, which might be feasible if it's just for one short-term project.
posted by bizzyb at 2:05 PM on March 11, 2011
Are you near a college campus? If so, you're probably able to use SPSS on campus in the library, which might be feasible if it's just for one short-term project.
posted by bizzyb at 2:05 PM on March 11, 2011
Sorry, you're right. It is by SAS, not by SPSS.
posted by Blazecock Pileon at 2:06 PM on March 11, 2011
posted by Blazecock Pileon at 2:06 PM on March 11, 2011
You can dload a free trial of SPSS at their website. Last I checked it was good for two weeks. Should be enough time to bang out an easy analysis like that.
posted by tiburon at 4:37 PM on March 11, 2011
posted by tiburon at 4:37 PM on March 11, 2011
If you (or your client) has excel, I'd start there. The data analysis ad in (I think that's what it's called) does everything you've listed.
posted by TheShadowKnows at 5:24 PM on March 11, 2011
posted by TheShadowKnows at 5:24 PM on March 11, 2011
Forgot about this one, but you can also download an R add-on that works through excel: http://en.wikipedia.org/wiki/RExcel
posted by bizzyb at 8:01 AM on March 12, 2011
posted by bizzyb at 8:01 AM on March 12, 2011
This thread is closed to new comments.
posted by djb at 10:30 AM on March 11, 2011