Minitab vs SAS In Six Sigma and Business Analytics

My new manager came from the marketing analysis group at our company over to the customer support group where I have worked for the last several years. His Master’s training is in econometrics which I learned is really just a fancy name for statistical modeling applied to economics. With that training and his previous positions he has used SAS as his tool of choice for statistical modeling and hypothesis testing. He obtained access for me to use one of our company’s SAS installations.

My background, on the other hand is in psychology, information systems, and Six Sigma. Thus, while studying psychology, I used both SPSS and SAS (one of my professors had a healthy research program on Behavioral Genetics and with his biology bent, was accustomed to using SAS rather than SPSS. Most other psychology researchers used SPSS.) When I attended training and became exposed to Six Sigma, most practicioners used Minitab. It seems fairly clear, in fact, that Minitab has made significant efforts to capture that part of the statistical analysis software market as evidenced by the fact that the two main products they offer are “Minitab Statistical Software” and “Quality Companion.”

I have, on more than one occasion, stumbled upon seeming religious devotion to one of these statistical analysis packages over the other. Such devotion often leads to debates, which in most cases consist of friendly banter between colleagues with different statistics backgrounds.

Which of these packages is best suited to Six Sigma projects? Which provides the best tools? And, which is most convenient to use?

I cannot offer definitive answers to these questions but I have a couple of thoughts that you might find useful. These are based on my recent experience with Minitab 14 and SAS 9 (Unix hosted, batch processing installation–not the desktop version with a GUI):

First I should note that the installation of SAS that I use at our company is a command line install on a Unix server. While I have used the stand-alone local computer version in the past, I don’t currently have access to such a license.

In cases where I need to produce results to a statistical question quickly, I typically use Minitab. The primary reason I do this is that I do not have to write any code–just import the data, specify the model in the appropriate dialogue box and I’m done. Additionally, for some tests (e.g., a simple regression model with a single input or independent variable) I can generate a quick plot that illustrates the result quite nicely.

In fact, I have found that Minitab’s graphing abilities are fairly mature and aesthetically impressive compared to those I’ve seen generated by SAS. This is important when presenting your project or findings to an audience that is not necessarily statistically inclined. In my experience, executives tend to follow my “story” better when I include graphs that illustrate particular points. It is often a great way to get a point across without having to explain all the gory details of a particular statistical test. For example, I prefer to show a regression plot with a quadratic factor because most people follow the main idea of a plot more readily than a regression equation, particularly when the model includes a quadratic factor.

On the other hand, I appreciate some of SAS’s capabilities as well. I recently learned a bit about ARIMA models and their value in forecasting. While Minitab produces graphical correlograms of the auto and partial autocorrelations, I prefer the output produced by SAS because it seemed more thorough and compact. SAS simply places all of the output in a simple text file and displays the correlograms with ASCII characters. In Minitab it became a little cumbersome to have to switch back and forth between the session and graph windows.

SAS also seems to have a much broader range of modeling procedures available compared to Minitab. If ARIMA is not what you need, SAS offers lots of other forecasting models that Minitab does not provide. Of course, you may have to purchase an add-on to use those procedures.

When hosted on a high end server, SAS seems to handle large data sets quite nicely. This may seem trivial but there are cases where handling large data sets becomes important. I once incurred the displeasure of our IT staff when I attempted to create a statistical control chart in Minitab with close to one million data points. After several hours of processing, with my poor CPU fan spinning at full speed, my laptop’s power supply stopped working and had to be replaced. I quickly learned the value of sampling. I also came to appreciate the raw processing power of SAS on Unix.

Price wise the two packages are worlds apart. Minitab sells for around $1000 for a single user license. I called SAS to obtain pricing for a similarly equipped SAS license and was surprised to find that the quote came in around $5000. Multi-user, server hosted licenses would come in much higher, but might prove economical if your company has a large number of users.

Overall, it seems that Minitab is a great tool because of its intuitive interface, great graphing features, and solid offering in terms of statistical tests and models. It is also FAR less expensive than SAS if you have relatively few analysts running statistical tests and models. On the other hand, SAS seems to offer a more comprehensive set of procedures for even the most obscure statistical methods and provides industrial strength processing power (when installed on powerful Unix servers.)

In the end, the “Mini” portion of Minitab should have been a tip–though it is no children’s toy by any means. It could reasonably be considered “Mini” when compared with SAS, however.