Basic stats program for Mac and Windows
February 13, 2008 7:49 AM   Subscribe

Looking for recommendations for a cross-platform biostatistics program for my lab. Desired features, program-specific questions, and...

Our lab does basic research, most experiments won't need anything more complex than a repeated measures ANOVA and in general t-tests will suffice. We have a full-time statistician in the department we call on for anything more complex. Currently, we are relying on an amalgam of stats programs and would like to find one we can settle on to purchase for common use.

Any program we choose must be capable of running on Windows and OS X. We would like to find something that is user-friendly. Strong preference for programs that are perpetually licensed rather than subscription-based. I'd also prefer a program that trusts me not to install it where I shouldn't, as opposed to products with tedious activation schemes: I don't want to deal with unnecessary hassles moving licenses between computers every time a student leaves or enters the lab. Academic discounts are a plus.

What I'm looking for are comments on first-hand experiences with different packages to help me make up my mind (I'm the one tasked with finding a good program). I need to know how easy programs are to use, ease of exporting data, how well features translate cross-platform, and errata regarding licensing types, whether academic discounts are available, activation issues, and the like.

Specific questions regarding likely candidates so far (to give you an idea what I'm looking at):

GraphPad InStat: Seems great, but how old is it? How well does it run on newer hardware and OS versions?
GraphPad Prism: Would also fill the need for a graphing program; however is it worth paying that much more than the InStat license? More specifically, is Prism getting updated more often than InStat?

StatSoft Statistica: I have been using the Windows version for some time, but how's the Mac version in comparison? Website makes it look like it's been basically abandoned for years...

Stata: While this one looks promising in terms of academic discount and Mac support, (a) it seems to be overkill for our purposes and (b) the command-line interface will not go over well with my mentor. In practice, how difficult is it for a non-statistician to use this kind of system? How annoying is the lack of copy-paste ability in the output window?

JMP: My mentor has used StatView for years, so JMP was a natural to try: We've both been frustrated by the interface and licensing issues though. Like Stata it's also overkill. While I wouldn't mind feedback on the program, we'd prefer not to go with this one if there are simpler, cheaper alternatives.

Free/Open source stats programs: There are several, from R down to a few actually developed by the stats department here at my university. If you've used them, or have one you particularly like, please feel free to add comments on how they compare with the commercial packages.
posted by caution live frogs to Science & Nature (10 answers total) 4 users marked this as a favorite
 
Stata has a command line, but really for simple things like T-tests you'll find everything you need in the menus (> Statistics > Linear models for ANOVA, for instance).
posted by methylsalicylate at 8:28 AM on February 13, 2008


Best answer: I find R and STATA terrifying.

GraphPad and SPSS are both user friendly and versatile. SPSS is more expensive and more fussy about installation, but I prefer it overall. I would download demos of both to machines at work and have your colleagues try them out and decide together.
posted by roofus at 8:39 AM on February 13, 2008


I really like JMP, but I really like the interface, so that doesn't help you. You might double-check with your university--we have a university-wide site license on JMP. Obviously, if you can get it free from your university that might sway things.

I've used R some and found it not that hard to learn, if you're at all comfortable with command-line interfaces. And it is free.
posted by hydropsyche at 9:15 AM on February 13, 2008


I tend to use JMP for normal stuff and R for when things get repetetive, weird or hard. You can get perpetual licences for JMP. I hate SPSS for many reasons, mostly their abusive attitude towards their customers. R is great but has some learning curve.
posted by singingfish at 12:48 PM on February 13, 2008


Best answer: I've used Graph Pad Prism on both Mac and Windows and I think it fits the bill for your requirements. The only problem with Prism is that it doesn't do Multiple regression (instat does, however), but otherwise it's great. Easy to use, and the results are in English. Eg. If you do a normal distribution test, it reports as such: Is the variable normally distributed, and answers Yes or no.
I use Prism for a lot of exploratory analyses before moving on to JMP for higher stuff.
Regarding Graphing, I find it quite adequate and highly customizable. The graphs are of better quality than Excel anyway.
And also, data entry for Prism is a bit different that in the standard big ones. For example, for ANOVA you need to enter the different variables in different columns as opposed to stacks.

(I'm in biology, btw)
posted by dhruva at 4:26 PM on February 13, 2008


I don't know how well it would suit your needs, but if they are truly simple, I use an open statistics program called PAST. You'll note that it's originally developed for paleontologists - because we're kind of "special", it has a very simple, easy to use interface. It's also completely free and extremely easy to install (I carry mine on a flashdrive), just uses converted Excel sheets to .dat files if you don't want to input data directly, and has a whole suite of regular statistics including chi-squared, F-test, two types of T-test, ANOVA, etc. The graphics could use lots and lots of work - they had good statistics in mind, not graphic design - but if you want to run a simple test for Linear Regression, then 2 clicks (Statisics - Linear regression) and you're done. Also, it says Windows only but I have zero trouble with it on a Mac. There aren't any licenses associated with it, either.

For what I do, in terms of ease and use, I like it a lot. You do have to have a good solid understanding of statistics to understand the answers. I usually have to reproduce any graphics, which is fine with me, but other then learning the right way to import data (you should play with the example files) it was very easy to learn. Of course, I could love it because I haven't really used anything else, haha. If you look into it, don't be fooled by the older looking website - website design comes and goes, but math is forever.
posted by barchan at 4:57 PM on February 13, 2008


SPSS is point-and-click and easy to learn. We use it in our Intro to Stats classes. You can dload a free trial. We get academic discounts from http://www.e-academy.com/
posted by tiburon at 7:47 AM on February 14, 2008


Response by poster: Thanks all - I had been leaning towards the GraphPad stuff, and am planning on throwing demos at my labmates this week for feedback. I'd be happy if Prism fits our needs (lack of multiple regression doesn't seem to be a problem for us).
posted by caution live frogs at 9:44 AM on February 14, 2008


I think that you should ask the consulting statistician. When something inevitably explodes or requires a more sophisticated analysis, you want to be able to send your data files / work to her directly. Similarly, she'll be able to send you code in the quickest time if it's her favored software.

R is pretty terrifying. However, some of the GUIs are fine (I particularly like JGR), and given the very limited number of tasks that you want to perform might be adequate. It's free to play with, so nothing to lose by checking it out.

Googling reveals some plugins (example) for excel that do repeated measures ANOVA; it might be worthwhile to look into excel plug-ins.
posted by a robot made out of meat at 11:04 AM on February 15, 2008


Response by poster: Just in case anyone finds this and uses it as a guide - I went with Prism and have been damn impressed. It's dead simple to use, produces great graphs, and is as user-friendly as you're ever going to find in a stats program.
posted by caution live frogs at 8:53 PM on April 30, 2008


« Older How can I repair my past mistakes without losing...   |   I need a Hero! Newer »
This thread is closed to new comments.