Completely random?

[Nucleus]

Ok now I need to generate a really large random number, between 9000000000000000000 and 9999999999999999999. Now the rnd function on its own doesn't generate a small enough number to be useful in the situation. I multiplied three rnd number together to get an extremely tiny number.

The question:

Have I introduced any bias?

Here is the function:


MsgBox Len(Random("9000000000000000000", "9999999999999999999"))

Private Function Random(ByVal Min$, ByVal Max$) As Variant
Dim x, y, z
Randomize
x = CDec(Min): y = CDec(Max): z = CDec(Rnd) * CDec(Rnd) * CDec(Rnd)
Random = CDec(((y - x + 1) * z + x))
If InStr(1, Random, ".") Then Random = Left$(Random, InStr(1, Random, ".") - 1)
End Function

[HarryW]

I have no idea if you've introduced bias. I can suggest a way to get a long number without multiplying though - concatenate random numbers in some range like 0 - 9999 and use a string to store it.

[Guv]

If your application is critical, I strongly advise some research in a local library or on the Web. Unless you are quite knowledgeable, you are likely to do something wrong.

Random numbers are far more subtle than you might suspect. There are all sorts of traps and problems in their use and much controversy about the definition of them.

In general, you must be aware that computer generated random numbers are referred to as pseudo random numbers by experts in this area of knowledge. This is to remind them that computers do not generate truly random numbers.

Furthermore, the random numbers suitable for one purpose are likely to be lousy for another purpose.

Instead of reinventing a wheel, do some research. A worthwhile pseudo random number generator almost always requires a fairly simple algorithm. I am sure that there is literature on this subject which would give you all the information you need to write your own generator.

You must choose the generator to match the process you are simulating. Do you need Poisson, binomial, or normal statistics. Perhaps you need something else.

To give you an idea of the possible pitfalls, are you aware that the VB Rnd function is a reasonably good, but not a precise generator for simulating craps? The random number generators built into previous versions of Basic (Quick Basic & earlier versions) were so bad at simulating craps that they indicated that the game favored the player.

There is an On-line service which uses radioactive decay to provide subscribers with random numbers generated via some radioactive decay process. I believe that this process generates Poisson statistics, which is great if the process you are simulating requires Poisson statistics, but questionable otherwise. Perhaps there are other similar services.

[jim mcnamara]

FWIW search in google for random number generator. The algorithm is dirt simple.

Guv is right - most rnd's are not really random - they don't pass the Runs Test for randomness. See this site

hth