This is not meant to be critical of Alan who has been active on this list
for probably as long as I, and he's certainly welcome to his own
preferences, but I couldn't help putting in my two cents.
Those of us who have been working seriously with computers from the days of
card readers and keypunch machines know that a so-called GUI is not simply
cosmetic, any more than the introduction of CRT terminals and, dare I say
it, personal computers and workstations are. (How many of us pine for the
days of JCL cards?) I think anything that makes software easier to use is
beneficial for no other reason than it makes users more productive, and it
seems to me that the bulk of the changes and additions to the latest release
of Stata fall squarely in that camp. This is not an argument in favor of a
GUI over a command-line interpreter, since Stata has both and we're free to
use either whenever we want. My point is that I feel Stata made a wise
decision to devote so many resources to increasing the flexibility and ease
of use of their software.
The suggestion seems to be that we would have many more analytical
procedures if resources hadn't been diverted, but I'm not sure this is the
proper way to view the situation. Stata is and has been one of the most
easily extensible statistical analysis packages available. (S-Plus is
another beauty, but I found the learning curve for Stata to be much less
steep than the one for S-Plus.) SMCL and the enhancements to the core
graphics routines were developed in this spirit. If there is a shortage of
specific procedures, then the user community is partly to blame because we
were given the tools to write these procedures ourselves. After all, the
ability to write our own procedures is why many of us use Stata rather than,
say, SPSS -- not that I personally would care to have written the hundreds
of ados that have been contributed. Rather, I personally can benefit from
having access to them. If we had to wait for Stata to supply every
analytical procedure, we would never have gotten as rich a set of procedures
as we enjoy today.
For my money (and my money unfortunately ran out around version 6), I would
advise Stata to continue on their present course of concentrating on
improving the set of tools they provide us to write useful ados, while
working on ados as resources permit. Note here that providing tools
includes incorporating state-of-the-art routines for estimation in the
built-in portion of Stata, enhancing the ability of programmers to extend
Stata (I'm still waiting for the ability to write true functions), "building
in" portions of ado code for computationally intensive tasks, etc.
One last remark, since this all started with complaints regarding -list-,
about changes to built-ins. I happen to agree with everyone who has argued
that the old default behavior of -list- is preferable to the new default
behavior. My reason, however, has little to do with performance
considerations. The issue to me is whether or not changes to the default
behavior of an existing command will break existing ado, or even do,
programs. Stata has been extremely conscientious about maintaining backward
compatibility with ado programs, but it appears that unadorned -list- has
been changed substantially and this may create problems with some users'
ados. As a very minor example, -xi- uses -list-, so -xi- will potentially
run slower now. The issue really isn't whether -list- is now better than
before, it's whether old programs will continue to work as before.
-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of FEIVESON, ALAN
H. (AL) (JSC-SD) (NASA)
Sent: Wednesday, February 26, 2003 10:57 AM
To: '[email protected]'
Subject: st: one last question [Re: list in Stata 8]
Phil, Buzz, Nick and others:
OK - OK - I respectfully admit I am in a small minority (of one?). I just
have one last question, then I'll be quiet about this topic:
What percent of Stata Corp.'s resources were spent on the development of
GUI, graphics, windowing, etc. for STata 8, as compared with the develpment
of improved statistical modelling, estimation, etc. routines? Same question
regarding SMCL in Stata 7.
Al Feiveson
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
