gThe usual way of dealing with
discordant data, often well
disguised, is to reject those which do
not agree, after which the remainder
will be found to agree very wellh
—
Paul Merrill