|
Steve,
You
are battling something you can't win. The threshold value which can cause
undetected overflow errors is > 2^(b-1)-1 where b is 32 for 32-bit
systems. The largest value you can reliably represent as an integer is
2,147,483,647. You can verify this with by looking at
>
.Machine$integer.max
[1] 2147483647
The
number of 2,147,483,648 could (would?) overflow and could go back to
-1.
A
decent intro reference for these kind of issues is Numerical Issues in
Statistical Computing for the Social Scientist by Altman, Gill and
McDonald.
--Matt
Dear All,
I am using SPLUS Professional Edition Version
6.0.3 Release 2 for Microsoft Windows XP.
I found that if I do:
>1000000*1500000
# I get:
[1]
NA
A workaround is to
do:
>1000000.0*1500000.0
[1] 1.5e+012
Is there a way I can correct for the above bug without putting
decimal places?
Thank
you for your attention.
PS. No
luck in getting SPLUS 6.1 in my university yet, as they are out of
licence!
Steve.
|