The problem in Excel isn?t technically bias as Francois Dufresne pointe=
d
out. The problem is auto or serial correlation.
Another way of looking at it is that the conditional probability of a v=
alue
given the previous value isn?t uniform, leading to a potential problem =
if
you are generating, for example, annual loss ratios, and you want next
years ratio to be independent of this years ratio. (More commonly, you =
do
want some sort of structure, but you would want to build it in explicit=
ly,
and not have any more than you build in.)
The Rand() function in Excel does not have to be ?reseeded? (and there =
is
no option to do so). Unlike APL, it will start over each time it is loa=
ded.
(If I recall correctly, the initial seed is based upon the system clock=
..)
Less well known is that you can generate identical sequences of random
values, useful for testing and other purposes. The data analysis add-in=
has
a random number generator. In addition to providing options for other t=
han
uniform (such as Normal, Poisson, Binomial), it accepts a seed as input=
, so
results are reproducible.
I haven?t encountered any problems using the Rand() function in Excel,
although we use Fortran for serious work, and use a double seed process=
to
avoid the potential for serial correlation.
I?ll echo Michale Dubin?s question. Has anyone run into a material prob=
lem
using Rand in Excel?
=
Visit the CAS Web Site at http://www.casact.org
===============================================
To subscribe or unsubscribe from CASNET:
Send an e-mail to caslists@lists.casact.org
Type in the body join casnet to subscribe
or leave casnet to unsubscribe.