Accuracy of "StanfordPhd" Comparison Table

[Bob Muenchen]

Hi All,

I hope this isn't an inappropriate question for this forum, but I was wondering about the accuracy of the Stata information in this comparison table by "Stanford PhD":
http://stanfordphd.com/Statistical_Software.html . Has anyone had a chance to study it?

Thanks,
Bob

[Bob Muenchen]

Yikes! Sorry about the multiple postings. I was getting an error message regarding the link so I thought it did not post.

[Richard Williams]

There are several such comparisons floating around, e.g.

http://brenocon.com/blog/2009/02/com...as-spss-stata/

http://www.ats.ucla.edu/stat/mult_pk...e_packages.htm

http://www.burns-stat.com/pages/Tuto...e_statpack.pdf

I used SPSS for 20+ years and now barely touch it. I greatly prefer Stata. If my life depended on it I might learn R or SAS but for now I am perfectly happy with Stata.

[Nick Cox]

This is one person's evaluation. I don't think there is an analyst worth talking about who doesn't develop more affection for one program rather than another, and a little arbitrarily, use one more than another given interests and background, etc. I am old-fashioned enough to want to know who this is, and more importantly exactly what experience they have under each heading. Conversely, I have never used SAS and I have not used SPSS this century, and only occasionally used R and MATLAB (that's its name!), so I could not compete with this person's breadth of experience.

On Stata

1. "Still, in terms of programming flexibility, Stata and SAS do not come even close to R or Matlab." Without knowing what flexibility means for this person it is difficult even to discuss this statement. If it's hiding the trivial observation that it's easiest to program well if you know a language well, then so be it. Stata's combination of Stata strict sense and Mata makes for an unusual but effective range of programming styles.

2. Most of the blanks in the table seem accurate. One exception is nonlinear regression. Stata is quite well endowed there.

3. The whole analysis seems to underplay the importance of data management. I'd say Stata scores well on this.

I'd underline the obvious but crucial sample selection problem here. Ask on a Stata forum what people think of Stata and you will get many people fond of Stata answering. So, you might weight answers by # posts and probably inversely!

[Clyde Schechter]

Another omission in the comparison table is that Stata does do path analysis (-sem-). I don't know when that table was written: SEM is new to Stata as of version 12, I think.

[Andrew Lover]

The right tool makes any job easier- each program is just optimized for very different things.

R is indeed very, very flexible- I'd consider that a bug and nowhere near a feature as error messages are beyond opaque. Stata syntax is regimented which is brilliant for ease and speed of coding, but if you need non-standard analyses or data types, it can make things difficult.

One example- the vast majority of Stata regressions allow robust errors which are a trivial option, but user-written commands may not have it, so you could be stuck. In R, there are multiple packages with many different flavors of robust errors which can be applied to any object, but it's a serious pain to sort through it all and implement the 'best' one.