Please see comments below.
Nick
[email protected]
Mak, Timothy
Pls see below:
Nick said:
I find it difficult to follow Timothy's scenario of a small business
manager who wants "non-statistician friendly software" that offers him
[sic] (e.g.) a choice between a PCA and a rotated factor analysis. That
implies a knowledge of multivariate analysis that I didn't know was
typical of that market, but I may "misunderestimate" it, to quote a
recent President.
I reply:
Of course this manager does not need to understand PCA or factor
analysis - and that's the point, but everyone can look at a graphical
representation of clustering in data. If you can explore data
graphically from various angles with a few clicks of a mouse, I think
that's a very marketable function in a piece of software.
NJC >>> I can believe it. I readily guess it's true. I don't see any
interesting implications for Stata, however. That some people sell and
others buy snake oil is true, but surely not worth much discussion?
Nick said:
More crucially, perhaps, I can't accept Tim's gross contrast between
graphics in Stata and graphics in R. Stata graphics currently has one
very big limitation in my view -- no serious support for contour or
perspective view of three-dimensional data, i.e. surfaces z relative to
two other variables x and y. That aside, I believe that Stata and R are
broadly comparable in their degree of graphics programmability. They are
based on very different architectures, but only a few people need care
about that if the results are similar.
I reply:
I had a hunch that Stata graphics is programmable. However, as Sergiy
pointed out, the language is undocumented. As there is I believe no
course on offer in Stata on graphics programming either, I don't really
know where one can start.
Another point is that the graphical language in Stata is again different
from Stata script language. So altogether, if you want to master Stata,
you need to master 3 languages. Expertise in usual programming in R,
however, would help you a long way in programming for R graphics.
Finally, although you may have already seen this, check out:
http://addictedtor.free.fr/graphiques/
It may be that all these can be reproduced in Stata, but my feeling is
that the code would be more complicated, but I would gladly be proved
wrong!
NJC >>> Goodness knows where Timothy picked up these impressions. They
are misleading when they are not wrong.
First, Sergiy in his post is alluding, I believe, to _some_ undocumented
features that he, Sergiy, has found useful, but that can't be discussed
further unless and until he explains what they are. That Sergiy is a
clever guy is obvious. That he can ferret out useful details and make
use of them is apparent. However, that is not the essential way to go
beyond official Stata. In fact, it is a very atypical way to proceed.
Dozens of new graph commands have been written by user-programmers based
entirely on reading the Graphics manual and looking at ado code. Also,
the graphics commands are not qualitatively different from the rest of
Stata's language. They are just the commands in the Graphics manual.
What Timothy may have in mind is that the low-level class stuff is not
documented. That's true. It was for a while an intention of StataCorp to
document that, but they changed their mind. But probably only about four
people in the world really want that stuff to be exposed, Sergiy, me in
some moods and about two others. Those people are frustrated at least
some of the time, but I'll assert that the developer time freed by not
documenting that stuff (I guess a few person-years) is likely to bear
fruit in statistical functionality that hundreds of researchers will
want very much. StataCorp like everyone else have finite resources to
throw at unlimited wants and that was a hard decision, in every sense.
No doubt a few more graphics commands like these might be written by
users if a NetCourse were to be announced. But that's really secondary.
It's just like the current myth, which some people like to believe (we
all love good excuses), that you really can't learn Mata, because there
is no NetCourse. Meanwhile, lots of users are happily writing Stata code
with Mata components. How did they ever manage it without a book or a
NetCourse? Well, sure, a NetCourse would make things easier, and a lot
of people would sign up. But in the mean time there are two manuals. You
just need to read them and experiment.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
