Statistics
Developer: StataCorp
Commercial: (Price see below)
Version: 9.2 (12 Jan 2007)
Release Date: 2006-10-30
Last Updated: 2007-02-06
Pros: Professional statistical software package; Choice between graphical and command-line interface; Batch analysis; Stable and robust; Produces state-of-the-art charts.
Cons: Command-line mode requires getting used to; Output data difficult to export; Too big for simple projects; Documentation very expensive.
Description
Stata is not the right tool for figuring out what is left of your research budget or for calculating the average age of your students' class. Stata is for serious statistics. It even looks quite serious, almost intimidating, with a command-line and text-based output window. In this respect it falls behind SPSS, the more popular and more accessible statistics package, but once you have learned that the command-line can actually be quite handy and that every command is also available through the good old menubar, you start loving this software, seriously.
The biggest advantage of Stata over comparable software
packages is that it is programmable. This, together with
the command-line issue, might sound terrible to those who
have steered away from text-based interfaces since the
invention of the Macintosh, but I can assure you that it is
firstly not that difficult (Stata's syntax is simple and
the software's help works great as a reference tool) and
secondly you do not want to live without it once you have
to deal with really large data sets. For example it allows
you to conduct batch analyses, running regressions with
varying sets of variables and cases, do all t-tests at
once, etc. Of course you can take this to another level and
actually write little programs which behave according to
the results they find (e.g. automatically run a certain
regression in the case of a certain p-value obtained from a
certain t-test), although this might require you to delve a
bit deeper into the software's command structure. How deep
you can go is partly dependent on the depth of your wallet,
statistically sepaking, as the documentation for Stata is
just ridiculously expensive (prices depend on the version
of Stata you want to buy and the documentation coming with
it). But do not be intimidated by the prices given on the
company's website (somewhere between $400-$800
educational), because if you purchase Stata through one of
the distributors you can get it without any documentation
for around $100 (educational - do not cite me on that!),
although you then better make sure that you will have
access to the documentation somewhere else.
But as I mentioned before, you can get most out of Stata
without having to type a single line by using the menubar.
Here is what's on offer: summary statistics, ANOVA, linear
and non-linear regression, time series, models for binary,
count, limited dependent variables, and panel data, methods
for nonparametric, multivariate, resampling and simulation,
survey, and survival analysis. Stata also provides
transforms and normality tests, model testing and
post-estimation support, maximization of user-specified
likelihood functions and tools for epidemiologists. It also
provides data-management capabilities, again with batch
functionality which is great for those working with large
surveys (for example you could apply a certain function to
all values of variables starting with "income"). Access to
Stata's graphical functions can be tricky through the
command-line (better use the menus here), but the output is
very professional and of high quality. It offers more
options and output graphs than Excel, but if you are into
funky-looking graphs you should look at something more
specialised such as Aabel.
One last thing: Stata does require a steeper learning
curve than its main competitor SPSS, but it is also more
powerful and efficient once you get there. And what is
maybe even more important for the Mac user; Stata's
developers have a firm commitment to get new versions out
for Windows and MacOS X simultaneously and with exactly the
same features and degree of stability (it is solid as a
rock, by the way). This is absolutely not true for SPSS,
whose developers have the tendency to publish buggy and
unfinished versions of their software for the Macintosh.
Useful Links
UCLA Stata Help Site
Wikipedia Entry
IDEAS: Statistical Software Components, Boston College Department of Economics