home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.lang.misc
- Path: sparky!uunet!think.com!ames!purdue!mentor.cc.purdue.edu!pop.stat.purdue.edu!hrubin
- From: hrubin@pop.stat.purdue.edu (Herman Rubin)
- Subject: Re: languages which allow the introduction of new operators
- Message-ID: <BxtMoK.Js8@mentor.cc.purdue.edu>
- Sender: news@mentor.cc.purdue.edu (USENET News)
- Organization: Purdue University Statistics Department
- References: <Bxr9vx.KBD@mentor.cc.purdue.edu> <Bxpr2J.JF7@mentor.cc.purdue.edu> <721862492@sheol.UUCP>
- Date: Mon, 16 Nov 1992 18:14:42 GMT
- Lines: 51
-
- In article <721862492@sheol.UUCP> throopw@sheol.UUCP (Wayne Throop) writes:
- >:: The fact that the language is conceptually
- >:: stack based does *not* mean that it has *any* stack at runtime!
- >: From: hrubin@mentor.cc.purdue.edu (Herman Rubin)
- >: Message-ID: <Bxr9vx.KBD@mentor.cc.purdue.edu>
- >: This is the case if the considerations for the optimizer were put in by
- >: the compiler writer. Computers are extremely fast sub-imbeciles. If we
- >: had a group of compiler people who would do their best to incorporate
- >: considerations which users see in their optimizers, and produce their
- >: optimizers so that these improvements can be quickly inserted, we may
- >: get reasonable results.
-
- >Yeah, right, it's all those lazy compiler people who just aren't doing
- >their best. Just like those lazy statisticians. If we had a group of
- >statisticians who would do their best to incorporate the statistical
- >considerations which users actually see into their statistical models,
- >and produce their models so that improvements can be quickly inserted,
- >we might avoid all the misuses of statistics by poor, hard-working,
- >long-suffering scientists in other fields.
-
- As a statistician, I have absolutely no right to tell the client what
- the model should be. Nor do I have a catalog of models from which the
- client can choose. I WILL question the client on the assumptions, and
- point out the consequences, and I MAY suggest that certain assumptions
- can be dropped or generalized without affecting the analysis.
-
- Now I may or may not be able to devise an algorithm to handle the data.
- It may even be that there is nowhere near enough data. It may be that
- all the methods I know of attacking the problem require too much computing
- to be useful. BUT I WILL NOT TELL THE CLIENT WHAT THE MODEL SHOULD BE,
- EVEN IF ASKED. Any client who wants me to even work at this must make
- me a co-investigator in the field.
-
- >: The semantic model of the language can, however, greatly inhibit the
- >: programmer.
-
- >True. Just like all those nit-picky details they try to teach in
- >statistics classes can inhibit a scientist's creativity with numbers.
-
- If my programming style will cause incorrect results to occur, it can
- be criticized. But, for example, nobody has claimed that the use of
- gotos WILL produce wrong answers. I am quite willing to admit that
- some of the methods I want to use in programming are dangerous. So
- is using a statistical method for an incorrect model, and this is
- done not just daily, and not even just hourly, but probably every
- second.
- --
- Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
- Phone: (317)494-6054
- hrubin@snap.stat.purdue.edu (Internet, bitnet)
- {purdue,pur-ee}!snap.stat!hrubin(UUCP)
-