The release of IBM SPSS Statistics 19.0 is in the offing, so I thought I taken my traditional opportunity to rant about the Java platform that SPSS moved to after version 15 and offer a few suggestions regarding what SPSS could do to make me want to switch to SPSS 19 when it is released. Variable Folders and cleaner, simpler cut/paste options would go a long way to get me past my intense dislike of the Java-based interface.
Note: IBM/SPSS has announced the features of SPSS Statistics 19.
Of course I understand that the development costs are lower when you write only one version of the software that runs on PCs, Macs and Linux systems (although it doesn’t seem like the price is lower). I also imagine it makes it easier to do certain kinds of integrations with other languages (although I don’t know which ones). And maybe academia has embraced Java and prefers the Java interface for SPSS. And I do appreciate that SPSS has made every effort to try to make the Java interface as friendly and as similar to what I’m used to seeing as possible.
But it isn’t enough. SPSS versions 16-18 do not have the same look and feel as all of the other programs I work with on my PC every day. SPSS is more sluggish, the spacing of the fonts look weird, and whenever I try to copy and paste data from SPSS to Excel (which I do hundreds of times a day) it takes five or six times as long because I have to unformat the data that has been transferred.
This isn’t the first time I’ve had this rant. It happens every year or so when I say, “Ok — I’m going to give this next version of SPSS a try.” I put away my half-broken-under-Windows-7 version of SPSS 15, load up SPSS 18 (despite the fact that I never upgrade, we keep buying updated licenses) and give it a try — only to slink back to SPSS 15 once I’ve had my fill of frustrating, non-value added experiences.
With that said, there are a couple of things IBM/SPSS could do to make me finally make the switch to SPSS Statistics 19.0 when it is released.
Better Variable Organization Tools (Folders)
I’ve asked about this before, and I’ll keep asking! Many of my data sets have hundreds of variables (multi-response questions can do that to you). Life would be so much easier if it were possible to organize these variables into folders. Folders that could be closed so that variables that I’m done with can be hidden.
The ability to color code variable variables would also be extremely helpful. Even better — let me add folders that I can color code.
This is much more preferable to what I have to do now, which is create “fake” variables with repeating x’s (“xxxxxxxxxxxxxxxxxxxxxxxxx”) in order to make it visually easy to identify different sections of variables. And what kind of solution is that?
Another useful capability I’d love to see in SPSS 19 would be the ability to have “aliases” of variables. There are times when I like to keep a copy of a variable with several different groups of variables (for example, if I’m doing segment analysis). Instead of having to pull the variable out of it’s “natural group” of variables, it would be nice to be able to easily create an alias which I could then store with the other variables.
Cleaner Cut/Paste Options
I would switch to SPSS right now if it were possible to copy tables and paste them into Excel without all of the Tablelook formatting. I know someone must have thought it would be a feature to copy the formatting along with the data. This might even make sense if you were copying the data into Word or Powerpoint. However, there is no reason I can think of as to why I would want to copy all of that formatting from SPSS into Microsoft Excel. It is absolutely a nuisance. I would much prefer if IBM SPSS Statistics 19 gave me a choice (perhaps somewhere in the options) to be able to copy data without any formatting — much like it used to do in SPSS 15.
Pre-Defined Banners in Custom Tables
I won’t say that I need this feature, but when I’m working with SPSS Custom Tables it sure would be useful to be able to pre-define sets of variables that I could easily drag to the top of my table in a group — without having to hunt them down in the list of variables one at a time.
I realize my requests are pretty much focused on my work-flow and how I will/would use IBM SPSS Statistics 19, but I can’t imagine I’m the only one with these issues. And please, IBM, if you really want to make me smile go the extra mile and bring back the OS-specific versions of SPSS!
I am a Win7-64, v18 user (used to use v15/v17).
I agree with the formatting stuff. I am having trouble visualizing what you mean by’varable folders’. Can you draw me a picture? Are you thinking something like a variable explorer (as in the old SPSS Dimensions)?
I would also add that the viewer has become extremely sluggish (perhaps this is a result of the new java interface??). It takes FOREVERto cut out or copy large tables from the viewer . It even takes a lot of time for the viewer to display. Much worse that in v15.
My most wanted capability would be syntax ‘alias’ just like ‘Set’ in VB for apps. I would really like to create a global syntax file that I can update over time rather than going through all of my syntax and changing file names and otehr parameters manually. Perhaps there is already a way to do this using scripting or Python prog blocks (do tell if you know about this), but I’d really love to see this built into the syntax method.
Lastly, the program itself is very sluggish at times and prone to crashing and other oddities (I frequently kill/restart the process). It also hogs memory and takes quite awhile to release it back to the OS after the ‘work’ is finished. Same goes for the sort (which has always been WAY slow in all versions). Sorting goes through 3 iterations, and it takes about 3-5 minutes to sort 2 million cases with about 150-200 variables. It takes around 24 hours to sort 90+ million cases with only about 10 variables! Granted, these numbers are big, but other programs/interfaces sort data much faster.
thanks for the blog.
David
Probably the best example of “variable folders” I can think of would be be equivalent to layer groups in Photoshop — basically, you could take a set of variables, assign them to a group and then make them all vanish so that only the group header were shown. My data files would be so much easier to manage if I could put all of my demographic variables into a “folder” or group labels “Demographics” — and even better, put the different kinds of demographic variables into sub folders.
It just speaks to the fact that SPSS 18 and (probably) SPSS 17 don’t really do much to help users with their workflow. It’s as if they think we’re just going to run a couple of analysis and put it away until next time. Whereas when I work with a dataset, I tend to have to work all the way through it from top to bottom!
I agree with all your points, David — which is why I’m continuing to use SPSS 15 and have refused to upgrade — even though the disks are sitting right next to me (alright, practically right next to me).
Hi Mark,
i absolutly agree with your format hints about spss. Also i’m annoyed about cut/paste between different applications (mostly excel). I’ve been using spss many times and your whishes for 19 are brilliant. I never thought about folder options, but it could make sense in front of multiple responses.
I’m not shure if it is a Java problem about sluggish spss. The gist in spss is using a strong syntax. For myself, i wish more options in using the spss syntax for daily workarounds, e.g. Davids global syntax file. I’m not amused about using VB more often i would.
I hope the guys at ibm / spss will attend to this post.
Greets,
Andreas
I work in a medium sized market research agency. We have five licensed users of SPSS and about a year ago we all made the decision to revert from v.18 back to v.15 – which is much much faster. I’m arguing with SPSS about whether or not we should continue to pay the yearly maintenance fee!
I can empathise with your experience of using the new Java based platform. V15 is still easily the most reliable to work with on a day-to-day basis. I am looking forward to exploring V19 though. V18 was a big improvement over the original Java based versions, which were very good at crashing but not for getting any analysis done!
In terms of variable organisation, have you used the use sets function? This essentially lets you put variables into folders and select which ones to show/hide. I find it very useful for very wide data sets. But, perhaps you are after greater flexibility than this?
i have developt sript spss tables (Now, Custome tables) in excel by vba excel. It just like as a generating syntax spss in VBA Excel. It is Very usefull for Large Tables.
As of this writing, stay away from SPSS 19:
http://www.spssforum.com/viewtopic.php?f=5&t=5581
In my work we decided to switch to PSPP instead of SPSS. So far we’ve found it a superb alternative. It has saved the company $$$$ and workes exactly like SPSS so everyone already knw how to use it. The only problem is that it doesn’t yet have all the advanced statistics but fortunately we haven’t needed them yet.
All of you are right! I’ve using spss since 1991 and let me tell you. After SPSS 14, the only good thing I perceive in newest versions is its appearance. I prefer to stick to versions 13 or 14. I don’t like to use custom tables very often, I prefer Global tables, Tables of Frequencies, Basic tables, etc in the menu. Sorry if I bother the IBM/SPSS people.
I just switched to 19 recently and am having lots of difficulties with copy and paste. I am trying to enter data into a new variable and then copy it and paste it all the way down, to about 12,000 cases or so. It took me forever to do this yesterday, and then when I got down to the bottom and right clicked to paste, I waited and waited and waited and it never pasted. I just got the hourglass symbol and after an hour I just had to shut it down. I never had this problem in 15.
What is the short cut to copy all the way down to the end? I never needed a short cut in 15. It just zipped right down. But even if I do figure out a short cut, how long is it going to take to paste??
Organizing variables: variable sets can go a long way to organize variables into groups. Also the views via names or labels, sorting by file position, alphabetically, or by measurement level all help. Also, using extension commands, you can easily generate macros for selected sets of variables based on many properties, including custom attributes, measurement level, patterns in names (all variables whose name includes “age”) etc.
For one of the commenters: FILE HANDLE. You can factor out references to specific locations and files. And programmability makes it easy to organize code into functional modules.
Interface speed: Version 20 has dramatically improved the speed of table rendering – back comparable with what it was in V15. Also a number of fixes to data copy and paste.
variable folders – see Variable Sets. I think it does exactly what you want… pity there is no syntax for this (but settings are saved with a file)