[18969] in s-news-athena
[S] Variance calculation
daemon@ATHENA.MIT.EDU (Douglas Bates)
Wed Aug 18 16:33:58 1999
To: s-news@wubios.wustl.edu
From: Douglas Bates <bates@stat.wisc.edu>
Date: 18 Aug 1999 15:26:27 -0500
Message-Id: <6rvhacltsc.fsf@verdi.stat.wisc.edu>
Under S-PLUS 3.4 on a SPARC/Solaris system the following occurs
> var(rep(0.182, 2)) # would expect 0
[1] 7.70372e-34
> var(rep(0.182, 3))
[1] 7.70372e-34
> var(rep(0.182, 4))
[1] 7.70372e-34
Under S-PLUS 5.1 on the same computer the answer is 0.
For most calculations that difference would not be important but it
happens that I am taking the logarithm of the variance and trying to
screen out the replicated points by trapping on logarithms of -Inf. I
guess I will have to work out an alternative method of doing this
calculation.
If you test that calculation on your system, could you tell me if it
produces a non-zero result. I would be particularly interested in the
results from S-PLUS 4.5 or S-PLUS 2000.
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
