« New Approaches in Single-Firm Event Studies |
| "Empirical Holy War" -- Eisenberg v. US Chamber of Commerce »
Before I finally commit to upgrading to Stata v.11 I'd appreciate any comments (pro or con) from anyone already using it. Major new versions of any software scare me a bit.
Posted by Michael Heise on 23 September 2009 at 10:03 AM in Current Affairs | Permalink
I make extensive of Mathematica and am a huge fan of the product. The new statistics capabilities in Version 7 such as GLM, Probit and Logit regressions mean that empiricists ought to take a fresh look.
Four places to see what can be done include http://reference.wolfram.com/mathematica/guide/Statistics.html, http://demonstrations.wolfram.com/search.html?query=statistics http://www.wolfram.com/solutions/Econometrics/ and http://www.wolfram.com/solutions/Statistics/
Seth Chandler |
27 October 2009 at 12:46 AM
I should preface this by saying that my experience with computers pre-dates the point-and-click era. I did actual programming in BASIC, COBAL, FORTRAN, SPSS, and (my favorite still) BMDP when I started work in computer stats. The idea of programming, iow, is not particularly off-putting.
I've pretty much switched over to R for most of my projects. That doesn't mean I use it the same way as I used STATA, however.
I don't use it for data entry or major data manipulation; I use spreadsheets for that then read the data in. And I don't use R base for too much of anything: I use Gary King's ZELIG for almost all the linear model stuff and other "libraries" for most other purposes. ZELIG simplifies the entire programming effort for linear models to next to nothing and other libraries allow you to re-cap SPSS, SAS, and STATA functions for most purposes without undue work.
As an R novice, I have found Robert Muenchen's R for SAS and SPSS Users very useful in pointing out R's differences from other stat apps and a good set of warnings against any general weirdness. Gelman and Hill's Data Analysis Using Regression and Multilevel/Hierarchical Models is pretty good on R and STATA too; they usually have programs for both languages set out for each technique. (I have only begun to work through the first part of G/H, so don't hold me to this.)
Why take on the additional burden, you might ask? Well, R isn't easy and it is so flexible that it can be a royal pain in the tush at times, but it has several real advantages. It's free and it isn't platform dependent. It has the best set of graphic routines extant, albeit that they are hard to work with in some incarnations. R has a useful and pretty sophisticated menu driven GUI, John Fox's R Commander, so I can just breath a sigh of relief and point and click for some jobs. People are adding to it all the time and it always has cutting edge techniques available. And, I believe I mentioned, it's free! I got tired of paying the stat app companies a fortune every time I turned around.
I still have STATA 8 and I use it fairly often. I also still use DATA DESK for many purposes. But I'm trying to go all R all the time. I might add that the students don't find it that intimidating (they only use the Commander) and they love it that they can load it up for free and use it on their own. My colleagues are a bit more reticent, but they are interested.
Tracy Lightcap |
30 September 2009 at 03:12 PM
I find the main benefit of Stata 11 to be in the improved programming environment. "Do" files now automatically have line numbers and color code comments and other features. Loops using features such as "foreach" are flagged.
Substantively, the new ability to do competing risk models could be very helpful. If one is studying time to an event, such as trial or settlement, there are always competing risks, such as pretrial dismissal.
Theodore Eisenberg |
28 September 2009 at 10:53 AM
Stata 11 crashes with a slightly higher frequency than Stata 10, however I am still happy. As for switching to R, you should only do that if you don't co-author papers (the vast majority of potential co-authors use Stata).
24 September 2009 at 03:10 PM
I have Stata 9 and it sounds as if Stata 11 is a huge upgrade from that earlier version. But I have also been throwing around the idea of using R. Does anyone have any opinions about the relative utility of Stata 11 versus R?
David Stras |
24 September 2009 at 08:35 AM
I went ahead and got Stata 11 and mainly for the graphics as Frank suggests. The new version can use the fonts on your system, making it easier to integrate graphs from Stata with the formatting in a paper. Also, I needed to get the "SE" flavor and figured as long as I was upgrading I might as well get the new version. If I did not have these two reasons, I probably would not have gotten Stata 11.
Now for my question . . . is anyone making good use of Mathematica? I've had a few occasions to use it, but nothing fancy. I was wondering if anyone was making creative uses of it in ELS circles.
Bob Lawless |
23 September 2009 at 08:05 PM
Excellent point, Frank. I'm coming to the upgrade decision from v.10.
23 September 2009 at 12:00 PM
Don't be scared. I've noticed a few annoying differences in the dropdown menus, where you have to look up the answers, but nothing different when you just type in commands. What was your last version of stata? I think the big upgrade/change for most purposes was in the graphics for stata 10. I'm not sure it's worth an upgrade from 10, but I never got 10, and it's worth an upgrade from 9
23 September 2009 at 11:02 AM
The comments to this entry are closed.