Automatic statistical analysis - why bother?
Af: Jesper Ramsgaard
During the ExSl project it has been a integrated part o make the issue of performing listening tests a non-issue. What normally takes time when performing listening test is the separate parts of experimental design, data-gathering, and last but not least statistical analysis.
In many ways we have gone a long way in our development of an automatic package for performing statical analysis on the more "standard" test types. There will always be limitations in regards to handling data from experiments automatically. Input from our novel free association method, being one of them. But when dealing with regular output, automatic routines has shown itself to be a time saver of dimensions.
Its a 'pick your weapon' type of situation when dealing with tools for statistical analysis, and some people will get almost religious about it. But one thing is certain: when moving beyond mere excel based analysis, you need some firepower. The main tool we use for statistical analysis is R. R is a open source statistical engine, that has proven to be a cheap, and very flexible solution. There are the usual drawbacks when dealing with complicated pieces of software, in that you need some basic introduction (or good programming skills) before you will be able to do anything at all. But still an good alternative to the more expensive solutions, and more flexible/unrestricted in many ways. A good example is how quickly new methods/routines are implemented. Usually you just search the open source database for new "packages" and there you go.
I will recommend that you sign up for the many forums that can help you on your way (the ´The R Project for Statistical Computing´ on Linkedin can be useful) if you would like to get started using R. You are also always welcome to get in personal contact if you need more specific sparring/discussion.
The main place to get your hands on R is here.