Counting inexact numbers

[Peter Cameron]

On Tue, Nov 04, 2003 at 09:52:36AM -0500, jomorga1 wrote:
> I haven't read the paper, but do disagree with the "not very 
> meaningful" part. If user's design has 3 whatevers, user may also like to know 
> if fewer whatevers is possible. The reported value 2/3 tells the story 
> exactly: user has 3, some other design(s) have 2. User now has the information 
> that could lead to checking out designs with 2 whatevers and compare to the 
> design at hand.
>
Agreed - but (a) this information may not be reliable and (b) the user
knows that his design has 3 whatevers, and a database search should
easily find him the design with 2 whatevers. So someone who really wants
to know is not disadvantaged. Moreover, the user is now certain that it
is 2 versus 3 and not 4 versus 6 (which is the essence of my objection
to rationals here: counts form a nominal scale!

Peter.