Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ncar!tank!paideia!vevea
From: vevea@paideia.uchicago.edu (Jack L. Vevea)
Newsgroups: comp.lang.c
Subject: Re: standard deviation
Message-ID: <2466@tank.uchicago.edu>
Date: 26 Mar 89 19:31:45 GMT
References: <9847@ihlpl.ATT.COM>
Sender: news@tank.uchicago.edu
Reply-To: vevea@paideia.uchicago.edu (Jack L. Vevea)
Distribution: usa
Organization: University of Chicago, Dept of Education
Lines: 40

In article <9847@ihlpl.ATT.COM> kat3@ihlpl.UUCP (Craig,R.J.) writes:
>------------------------- cut here -------------------------
>The input data is given below:
>------------------------- cut here -------------------------
>10
>111111111.0
>111111112.0
>111111111.0
>111111112.0
>111111111.0
>111111112.0
>111111111.0
>111111112.0
>111111111.0
>111111112.0
>-------------------------- cut here ------------------------
>
>
>Notice that the onepass standard deviation is 1.78, while the twopass
>standard deviation is 0.5!!!!!!  Incidentally, the true value is 0.5!!
>


	Sure; and you could just as easily have picked another data
set that produced a correct result from the onepass program, and an
incorrect result from twopass.  Which is why, as has been pointed out
before, neither algorithm should be used in a general standard
deviation program.

	There was a posting recently to sci.math.stat, not crossposted
here, that gave a reference for better computational algorithms.








Saepe fidelis.