This is something that may interest precisely zero members of our reading audience. As it is, it only really interests 50% of the authors.
When you learn econometrics or other types of statistics as a graduate student, you normally end up having one of two things happen to you. In the one instance, you learn one software package very well to fit the types of research questions that you like to ask. In the other instance, you learn a whole bunch of software packages to reflect the fact that you have eclectic research interests. The problem, as you might imagine, is that some statistical packages are extremely good at doing some things, but almost none are good at doing everything. I (TP) have encounted this problem often. If I want to study time series cross sectional data, Stata is the way to go. For survey data, SPSS is king. But one at least one occasion I have been faced with a problem of generalized autoregressive conditional heteroskedasticity (GARCH), which is almost impossible without a software package normally used for financial analysis called EViews. I have to learn to use all of them. It is not unheard of for people to tailor their research to the types of questions that their preferred statistical software can answer.
None of this would be a real problem if statistical software were free or easy to learn to use. But neither of those are true. Which is why it is particularly unfortunate that no one in my graduate education encouraged me to learn R. After four years, I have finally decided to make the switch to R, which is not only extremely powerful, but FREE. It also, like good open-source software, does not take up much hard drive space. It’s a little challenging to learn at first, mainly because it is command oriented (unlike other programs that spoil you with fancy pull down menus). But it is no harder to use than the other very powerful command oriented statistical program, Gauss, which I had to learn, and which is quite expensive if you want the version that handles lots of data.
And also, as is the case for other open-source software, there is a huge community of R users who write libraries for all sorts of estimators and problems. It took me 2 seconds to find a GARCH library that is just as easy to use as EViews. When I am a professor, my students will be contributing to the revolution by using R. If you are one of our readers who struggles with econometrics software–and I know that you are out there–may I strongly urge you to look into R. Imagine, never having to pay for stats software again.
(Shhh, but it also has the added benefit of making you look smarter simply because you use it. Using Gauss does the same, but R is free. This much like composing your papers in LaTeX instead of Microsoft Word, which I am almost positive increases your chances of publication.)
Matt Glassman July 19, 2005
Tom – I have been trying to learn R for the better part of a year. I mostly like its graphic capabilities. I have found it thoroughly frustrating, mostly because there aren’t many great starter manuals – they are either too small or too comprehensive. Any pointers to good ones?
Tom July 19, 2005
There’s the one at CRAN (http://www.r-project.org/manuals.html), which is OK, but the best deal is the one that you get with Zelig, Gary King’s attempt to make us all use R. It’s at gking.harvard.edu/zelig. Twenty cents if you get what the name is a reference to.
Basically, with this, you get instant functionality for most models that you’d want to run, plus all the best post-estimation commands. The documentation is great.
The hardest thing about R is reading in data, something which I have yet to start working on. But then again, my dissertation involves no regressions, so I may not work on that for quite some time.
What types of thing are you having trouble with?
Dave July 24, 2005
The big payoff of this otherwise geeky post is reading the word LaTeX then giggling. tee hee hee. Oh wait, then the uber-geek in me remembers that it’s pronounced Lah-Tech. sniff