I am getting involved with several clinical research projects and am looking to learn how to do some statistics. Anyone have any suggestions for which statistical analysis software would be the best/easiest to learn to use? I'm good at math/statistical concepts, but have little experience with writing computer code and I'd like to avoid coding as much as possible. I'm trying to decide between SAS, SPSS, and Stata, but if anyone has other recommendations, I'd love to hear about that as well. Thanks!
It depends on what you want to do. For beginners, I would go with MiniTab, you can download a free 30 day trial from
www.minitab.com. SPSS also has a trial, but runs out after 15 days (or 15 uses...something like that). Ultimately, in terms of publishing papers, SAS is the way to go, and at the VERY least, I've seen SPSS appear every now and then. Therefore you should work towards becoming relatively proficient with SAS.
There are three weaknesses (from my limited experience) with MiniTab:
1) One weakness with MiniTab would be the way it runs ANOVA. Its data output isn't as nice as SPSS'. So when I run an ANOVA, I usually use SPSS. But for the typical t-tests, bland-altman plots, non-parametric statistics (Mann-Whitney, Kruskal Wallis, etc), I just run it on MiniTab for quick results.
2) The other issue would be multiple pairwise comparisons. It doesn't have a great deal of options such as the Bonferroni-Dunn which is more conservative in my opinion and thus a better way to compare results among the groups tested via ANOVA or Kruskal-Wallis. Howover it does have the Tukey's so that is a decent substitute.
3) The way it calculates P-values is another issue. Its not serious, but sometimes it may report P = 0.000, which one should NEVER report
. SAS does a good job at giving a actual value. Of course you can always report the P = 0.000 as P < 0.001.
In regards to say logistic regression, I would REALLY go with SAS. Its a PAIN in the beginning to learn how to do it, but a quick google search can yield the lines of code that you would need to run such analyses. MiniTab and I assume SPSS also have the ability to do logistic regression, but for the most part, when I ran this type of regression, I went with SAS.
Now in terms of cost, SAS is probably the most expensive, and MiniTab is the cheapest. You can get a lifetime license for MiniTab for $100, or a 5 month license for some smaller amount. I think SAS is only a year long license. I have not looked into getting SPSS.
So back to my first statement: what do you want to do? If you're doing simple comparisons of means/medians, then SPSS or Minitab may suffice. SAS may be overkill. But if you're doing logistic regression, poisson regression, etc, then yea you'll need SAS for that. Additionally if you want to do some kind of epidemiological study, then I think SPSS is good too. Our MPH students seem to stick with SPSS more.
One last thing to consider is what do people use the most? Beginners (e.g., students, some researchers) use MiniTab and SPSS. However I can guarantee that biostatisticians primarily use SAS. So if you ever consult with a biostatistician, then it may be helpful to show them your SAS readouts, and they can help you with the results or lines of code if anything went wrong. Hope that helps!