For those of you out there who have used various statistical packages, what do you like best? I'm currently enrolled in a math class that requires the use of different packages for different projects.

What are the strengths of the ones you have used? Weaknesses? Which ones are user-friendly and easy to learn? Which ones are the most powerful and can perform the most detailed analysis?

Up to this point, I have used SAS (although not much), SPSS (nice for backward regression) and R (still learning this one- the syntax is confusing but it seems powerful). Others that I've heard mentioned are S+ and Jump, although I'm not familiar with them. I've also used Matlab for simulation exercises, but unfortunately Matlab is not very detailed when it comes to statistical analysis.

I'm asking here because my fellow students here either have detailed knowledge of these programs but aren't able to explain them or they are in the same boat.

Only used SAS so far. U should try 2 get some elaborated xamples 4 inspiration, it's very powerfull, but it sux ass 2 learn/work with (in my rather limited xperience).

The best option depends on the statistical packages involved as well as on your preferred operating system. The PC-based program Stat/Transfer (available on netware) can convert the following files -- SPSS, SAS, STATA, S-Plus, SYSTAT, Excel, dBase, Gauss, among others. The UNIX utility SAS2STATA converts files from SAS to STATA but is available only on UNIX. Some software packages can directly read in data sets from other packages. In addition to the options outlined below, one always has the "brute force" option of writing out data in ascii format in the original package and then reading in the ascii data in the new package.

SpcMs said:

SAS has given me SEVERE issues with saving data tables. Also, I haven't been able to figure out who to do backward regression with interaction terms which eliminates the interaction terms BEFORE the variables themselves, regardless of significance levels. There are manuals here at the Math Library, but they don't seem very helpful for that. A lot of my classmates have used SAS extensively and STILL have the same issues as I have had. Unfortunately, it seems SAS has become the favorite of modern statisticians, despite its difficulties.

"Hi, I'm from UPS, and boy do I have a package for you!"

2the9s said:

Thanks! I think I'll check out Stat/Transfer. Do you know if it is compatible with R?

tommyalma said:

I suppose I should change the title to "Statistical Software."

Moonbeam said:

You mean that made sense?

I just Googled that up.

spss...
used it for the last few yrs and is pretty good and easy to use...

2the9s said:

2the9s said:

Yes it made sense.

Anyone else?

Moonbeam said:

SPSS-10 is all that I will ever need for my type of statistics that I need to do etc..it does the regression, factor analysis, regression to the mean etc...and ANOVA and ANCOVA...effect sizes, r9Corr) and R...all with the help of excell mind you.



.
[This message was edited Tue Jan 21 22:05:31 PST 2003 by rdhull]

rdhull said:

I like SPSS. The only thing is that it is a pain to eliminate variables one by one when you have specifics. SPSS wanted to eliminate one of the 3 variables and leave in interaction terms involving the variable it deleted!

Moonbeam said:

Well that screws up the whole design then..how can the AxB , interaction, still be involved if one of the variables that is included in the AxB is missing...skewed etc...plus I think they are about to come out with version 10 ( or is it 112, 12 now?) that is updated etc..little by little they improve it with all these negatives things mentioned etc fixed.

rdhull said:

Exactly. That's why I had to opt for R on this project- it made things much easier to do by hand. Rather than making different data sets, I just had to update the model within R itself and voila!