Este es una respuesta que le di a una profesional que postea en Methodspace (www.methodspace.com), pero que me parece puede ser de utilidad para todos. Pongo la respuesta tal como la escribí, pero después la traduciré, espero le sirva a alguien que esté con la misma duda.
Honestly I think PSPP is far from SPSS and now it is a little bit outdated. I used PSPP some years ago, but now I’m using R, however I know it has a very steep curve of learning, so maybe it is not an alternative. Maybe a better alternative for you is OpenStat (http://statpages.org/miller/openstat/), however it has some bugs and it never works as it should be.
Other alternative is SOFA (http://www.sofastatistics.com/home.php) I still didn’t use it, it’s quite basic, but maybe for you will work.
Other alternative that works for me is TANAGRA (http://chirouble.univ-lyon2.fr/~ricco/tanagra/en/tanagra.html) , it is a data mining program indeed, but honestly, I can’t understand clearly the difference between general statistic procedures and data mining statistic procedures. Honestly, i think there isn’t any difference, exception made with some learning algorithms developed specially for to get inductive information from a huge volume of data. It’s easy to use, it’s very complete and it’s beautiful.
My last suggestion is PAST (http://folk.uio.no/ohammer/past/). It’s a great package, in spite is created with paleontologist in mind, it’s a very general statistic package. It’s small, but extremely powerful, it has many multivariate routines than you can find only in commercial packages. It has some drawbacks: a) it cannot manage ommited responses and b) it has a different system for to selecting variables and processing them.
My choice is TANAGRA, but you can decide for yourself.
PD: But if you accept advices, go for R… it will be the future…