Round numbers for parameters
Posted: Tue May 27, 2008 8:20 pm
Has anyone else noticed that round numbers for parameters often test worse than surrounding numbers? IE 19 or 21 tests better than 20, 99 or 101 tests better than 100.
Let A = (the set of round numbers between 0 and 10,000)
Let B = (the set of not-round numbers between 0 and 10,000)
Which set is larger, A or B?
Estimate the ratio R = (number of members of A) / (number of members of B).
An item "X" is chosen at random. X is either a member of A, or a member of B. Which is more likely, X is a member of A ("X is a round number") or X is a member of B ("X is a not-round number")?
Estimate the probability that a randomly chosen item "X" will be a round number.