>>Is anyone aware of reliability checks of FoxPro's RAND() function to determine whether it adequately
>>simulates random numbers on the interval (0,1)? Thanks.
>
>Superficially, it looks pretty decent. I ran the code below and got the following results for 1M iterations:
>0.0 < 0.1 99,986
>0.1 < 0.2 100,275
>0.2 < 0.3 100,187
>0.3 < 0.4 99,832
>0.4 < 0.5 100,005
>0.5 < 0.6 100,122
>0.6 < 0.7 100,162
>0.7 < 0.8 99,847
>0.8 < 0.9 99,815
>0.9 <= 1.0 99,769
Danged PRE - here's a more readable post:
Yes, RAND() is good. I did some tests using SAS a couple years ago, and vfp RAND() came up a statistically very random Uniform distribution. I use it in statistical work without concern here.
Meanwhile, ironically SAS failed miserably with its equivalent Uniform random generator (and it's a statistical app). Pascal also had a bogus one, it would give the same second number each time after a restart, something like that :)
We have a professional procedeure developed by a bunch of statisticians here using prime numbers to generate random uniforms from mainframe days, but the vfp RAND() is probably just as good as that.
The Anonymous Bureaucrat,
and frankly, quite content not to be
a member of either major US political party.