My well worn 4 year old Compaq desktop, from which I do most of my statistical analysis, is finally getting slow enough that I need to get a new computer. The rest of the computers, phones, and tablets in my house are Apple product, so I am looking at getting a Mac Mini. I'll be reusing all the rest of the peripherals from the Compaq for now as they work fine and I am not hung up on getting their Mac equivalents. However, this switch in computer hardware and thus OS is presenting an interesting opportunity for me.

I have an older version of Minitab that I run for most back end analysis for this blog. I then export Minitab's results to Excel 2010 to make the graphical presentation of the data look a whole lot better. I've learned to use Minitab via my Six Sigma training, and it's very useful for what I do. The trouble is, I only have a Windows version of it and would really prefer to not have to constantly switch over to the Windows OS via Boot Camp on the Mini whenever I want to do analysis and/or blog. It kind of defeats the purpose of getting the Mini. I'd like to do the analysis and blogging while running Mac OS. Here's the trouble - a Minitab license for a Mac does not exists, so purchasing an update for my Mac is not possible.

Luckily, I can get a steeply discounted (and legal!) copy of SPSS for Mac OS or Windows. We're talking less than $100 for a package that retails for $2700! The package available to me comes with the following SPSS modules:

- SPSS Statistics Base
- Advanced Statistics
- Custom Tables
- Forecasting
- Regression
- Tables Original
- Trends Original

My concern is at a minimum maintaining the availability of analysis in Minitab. Losing functionality I use routinely today is not an attractive thought to have My most frequently used functions in Minitab are:

- Descriptive statistics: normality plot and check, mean, median, standard deviation, quartiles, etc.
- Normal Sample comparisons: two and one sample t and p
- Correlation tests
- Box-Cox Transforms
- Non-parametric tests such as Mann-Whitney
- Linear and Non-Linear Regression with ability to generate prediction (PI) and confidence interval (CI) data off of "new observations"
- Binary Logistic Regression (with PI and CI data generation capabilities)
- Ordinal Logistic Regression (with PI and CI data generation capabilities)

**Does anyone know if the SPSS modules I listed contain features (4) and (5)?**

Finally, beyond the actually calculations available, I am interested in understanding the user interface for SPSS. My father used SPSS 30 years ago when completing his operations research masters. It has a good bit of legacy code associated with it, much like Minitab. That can sometimes limit the GUI overlay capabilities - Minitab can have some frustrating limitations due the legacy code. Having no experience with SPSS, I am concerned I might run in to different limitations.

**Does anyone know if SPSS's**

**UI is as good as Minitab's or better?**

Thank you to anyone who does provide feedback, and let me know of any other pitfalls or benefits I may have missed.

Stata would be a better option than SPSS, although it might take some time to get to know how to use it.

ReplyDeleteIf you prefer GUI based packages, you should definitely take a look at GRETL. http://gretl.sourceforge.net/

There is a native package for MacOSX, it is free, very intuitive and it seems to fit excellently your needs.

If you are ready for a greater change, then I'd recommend GNU R, which is a powerful and free language for statistical computations, with practically endless features and beautiful graphs. However, the learning curve is quite steep.

th.alys -

ReplyDeleteThanks for the feedback. I downloaded GRETL tonight, and it did seem very straightforward. The only challenge I had was in attempting to plot prediction intervals. It never seemed to be able to do it, even though it appeared I was requesting it (Analysis -> Forcasts from a Binary Logit model). Any insights as to how I might be able to generate such graphs/data as it is key to my blog. Thanks again for all of your help so far!

I'm not sure what's going wrong. Take a look here http://www.learneconometrics.com/gretl/using_gretl_for_POE4.pdf

ReplyDeleteWhy not use R instead of SPSS? Does everything you want and more, has very good graphics output (with some work required),is free, and is growing fast. Vast amounts of help, too

ReplyDelete