qr_points.txt
18 points to fit into one paper
--------------------------------------------------------------------------------

A) Quasirandom numbers are related to random numbers, but have important
   differences.

B) Random numbers are useful for sampling.

B) A simple sampling problem estimates the area of an irregular region.

D) The convergence rate for quasirandom sampling involves the product of
   two factors, which suggest it can only beat pseudorandom sampling for 
   relatively small dimensions (say less than 10).

E) Sampling with quasirandom numbers worked better on the CMO problem than
   other sampling methods.

F) Problems involving high dimensional variables become extremely difficult
   to solve, even with sampling; this difficulty is called the curse
   of dimensionality.

G) Random numbers are unpredictable, uncorrelated, and unbiased.  
   Pseudorandon numbers are predictable, but uncorrelated and unbiased.
   Quasirandom numbers are predictable and correlated, but unbiased.

H) The correct term for random numbers on a computer is pseudorandom.

I) The Monte Carlo method has a convergence rate that is independent
   of the spatial dimension.

J) Discrepancy is a measure of uniformity of distribution, that is, of
   how unbiased a set of number is.

K) Quasirandom numbers have higher uniformity than random or pseudorandom
   numbers, and this means they do a better job of sampling.

L) A difficult sampling problem involves estimating the current value of a 
   collateralized mortgage obligation (CMO) over 360 monthly payments.

M) The Monte Carlo method is based on sampling; the sampling can be done
   using pseudorandom or quasirandom numbers, or an orderly grid.

N) Quasirandom numbers can be generated by a formula.

O) Quasirandom sampling sometimes produces good results despite the curse of
   dimensionality.

P) We can think of sets of numbers as truly random, pseudorandom,
   quasirandom, or orderly.

Q) Quasirandom numbers are more uniformly distributed than pseudorandom numbers.

R) The CMO results, for dimension 360, suggest that quasirandom sampling can
   beat pseudorandom sampling in a high dimension.  This may mean that the
   actual problem being studied has some underlying simple structure that
   makes this possible.