Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83 based; site hou2b.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!lll-crg!dual!qantel!ihnp4!houxm!hou2b!halle
From: halle@hou2b.UUCP (J.HALLE)
Newsgroups: net.med
Subject: Re: The Perils of Nutrasweet: digits of precision
Message-ID: <592@hou2b.UUCP>
Date: Tue, 13-Aug-85 14:42:03 EDT
Article-I.D.: hou2b.592
Posted: Tue Aug 13 14:42:03 1985
Date-Received: Sat, 17-Aug-85 06:45:42 EDT
References: <587@rtech.UUCP>
Organization: AT&T Bell Labs, Holmdel NJ
Lines: 14

Before this debate gets too out of hand, it might be prudent to remind
people what the terms mean.  Accuracy and precision are NOT synonymous.
Precision refers to the number of significant figures, but is not
necessarily a good indicator of accuracy.  Accuracy refers to the
reliability of the stated value.  E.g., you could measure something
to be 13.246792 inches long, but if your ruler was only so-so and/or
your measurements had a lot of scatter, you could be accurate only to
perhaps two decimal places.  (13.246792 +/- 0.05, e.g.)
It is not necessarily incorrect to state it this way (although it is
bad form).  0.007% has three decimal points worth of precision, but
that merely tells you the size of the scale.  Its accuracy is unknown,
though implied.  Depending on the measurement method, three decimal places
of precision could be "significant," or it could be a crude estimate.
More information is required.