Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: GLM and ANOVA complaints

From   "Nick Cox" <[email protected]>
To   <[email protected]>
Subject   RE: st: GLM and ANOVA complaints
Date   Mon, 29 Sep 2003 11:46:04 +0100

Dave Airey wrote, among other things,

> The fact that many rely on desmat and its authors generous efforts
> means it should have been duplicated by Stata or incorporated as an
> official command by now.

The fact here is undoubted, but the principle Dave invokes is

-desmat- is one of many user-written programs which are very much
respected and very well used. However, it doesn't follow
inevitably that it should be adopted by Stata Corp.

In saying this I am not writing against -desmat-, which
as it happens I almost never use, not that anything whatsover
can be inferred from that. (I don't use -xi- that much
either.) I just want to put that comment by Dave in
a much wider context.

Among the issues which affect adoption of user-written software by
Stata Corp are

0. Whether it's important or indeed worthwhile. User demand
is naturally one key source of information here. In addition,
Stata Corp reserve the right to have excellent ideas and then
wait for the users to see how good they are. They also
reserve the right to be a little bit capricious.

1. The need to test, maintain (fix bugs; update as other
aspects of Stata change) and where appropriate
enhance software.

2. The need to provide technical support and
documentation. Almost everybody has some concerns about
bloated documentation while at the same time wanting Stata
to expand to do twenty extra things! Everybody will want Stata
technical support at a very high level. Things being what they are,
there are some very popular user-written programs in which
only the authors (and possibly some other users)
could provide adequate technical support. That definitely
inhibits adoption.

3. Compatibility with the rest of Stata. An idiosyncratic
program, even if important and widely used, has much more
difficulty getting adopted than something smaller, and
perhaps less widely used, or less widely useful, that
fills a definite gap in Stata or extends it painlessly.

4. Avoiding short-termism. There is no attraction
in adopting something which will itself be superseded
by a later more general or deeper innovation. Of course,
users cannot see as far as even the next release. (Stata Corp
at this time probably can't all the way either.) Some things
remain on long-term agendas because no one at Stata Corp
has yet had the bright idea which will unlock the door -- just
as many of us have research problems which interest us, but no
glimmer of a real solution yet. Of course, Stata can drop
things as well as add them, but it's reluctant to do
that. (For example, Stata Corp showed no interest in
adopting what was probably the favourite of my own
programs, with very good reason, because Stata 8's new
graphics superseded it almost completely.)

5. Finite resources. As Rich Goldstein hinted,
prioritising something implies pushing something else
down the queue.

I'm a amateur programmer myself; I nevertheless imagine
that Stata users who are not professional programmers
(that means, amateur programmers and non-programmers)
often see adoption of user-written software as essentially
a formalisation of just four facts:

	the program exists and appears in good condition
	the help file exists, ditto
	it's popular, at least for certain groups
	Stata Corp approve in some sense

If these were the criteria, then probably a couple
of hundred user-written Stata programs qualify
immediately! For professional programmers, however,
getting the code to work and a help file up and
ready is often the least onerous part of the work.

As said, I want to generalise beyond -desmat-, but
I'll make one further guess in that case.

I am pretty clear that the popularity of -desmat-
is read at Stata Corp as a strong sign that many
users find -xi- inadequate. However, the Stata
inclination is to treat that as a signal to take
a very serious look at -xi-. Does it need extending
or rewriting? I have no idea whether anyone is
working on that right now, however.

[email protected]

*   For searches and help try:
*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index