Generating semi random numbers

[Matt Krevs]

Hi all

I'm doing some load testing and want to generate some numbers within a given
range. My problem is that I want most of the numbers to be close to the
middle, and only some at either high or low ends of the spectrum

eg if the range is 0-100, I want 70% of the numbers to be betwee 40 and 60.

From my vague memories from school maths, I kind of want to implement a bell
curve and use various standard deviation functionality. From my description
you can probably figire out I have fairly hazy recollections

Would someone be kind enough to direct me to a link that could give me more
information and/or some examples?

Thanks

 

[Matt Krevs]

Ahh.

Found a couple of posts to refresh my memory
http://forum.java.sun.com/thread.jspa?threadID=624504&start=0&tstart=0
http://en.wikipedia.org/wiki/Normal_distribution

 

[Boris Stumm]

I'm doing some load testing and want to generate some numbers within a
given range. My problem is that I want most of the numbers to be close to
the middle, and only some at either high or low ends of the spectrum

eg if the range is 0-100, I want 70% of the numbers to be betwee 40 and
60.

java.util.Random#nextGaussian() maybe?

 

[Daniel Dyer]

Hi all

I'm doing some load testing and want to generate some numbers within a
given
range. My problem is that I want most of the numbers to be close to the
middle, and only some at either high or low ends of the spectrum

eg if the range is 0-100, I want 70% of the numbers to be betwee 40 and
60.

From my vague memories from school maths, I kind of want to implement a
bell
curve and use various standard deviation functionality. From my
description
you can probably figire out I have fairly hazy recollections

Would someone be kind enough to direct me to a link that could give me
more
information and/or some examples?

Thanks

As Boris suggests, use the nextGaussian method of java.util.Random. This
gives you a distribution with a mean of zero and a standard deviation of
one. If you need to adjust the distribution, multiply by the required
standard deviation and add the required mean:

Random rng = new Random();
double value = rng.nextGuassian() * standardDeviation + mean;

Dan.

 

[Simon]

As Boris suggests, use the nextGaussian method of java.util.Random.
This gives you a distribution with a mean of zero and a standard
deviation of one. If you need to adjust the distribution, multiply by
the required standard deviation and add the required mean:

Random rng = new Random();
double value = rng.nextGuassian() * standardDeviation + mean;

Note however, that this will not guarantee the result to be in any fixed
interval, like, e.g. 0-100, as the OP requested. Maybe you can specify your
requirements more precisely. The binomial distribution could be a good choice.
Several distributions are implemented in this library:

http://dsd.lbl.gov/~hoschek/colt/

Cheers,
Simon

 

[iandjmsmith]

Note however, that this will not guarantee the result to be in any fixed
interval, like, e.g. 0-100, as the OP requested. Maybe you can specify your
requirements more precisely. The binomial distribution could be a good choice.
Several distributions are implemented in this library:

http://dsd.lbl.gov/~hoschek/colt/

Cheers,
Simon

Before anyone rushes to use this, perhaps you could check the problems
discussed in this thread
http://groups.google.co.uk/group/comp.lang.java.programmer/browse_frm/thread/dffff3adce0b23f2


If the problems have been fixed or did not exist in the first place
then fine. Otherwise I think it needs some maintenance work done on it
before use.

Ian Smith

 

[Simon]

Before anyone rushes to use this, perhaps you could check the problems
discussed in this thread
http://groups.google.co.uk/group/comp.lang.java.programmer/browse_frm/thread/dffff3adce0b23f2

Ooops, I didn't know that, thank you for this link. Then maybe one should rather
use the library mentioned in this other thread.

Cheers,
Simon

 

[Luc The Perverse]

Hi all

I'm doing some load testing and want to generate some numbers within a
given range. My problem is that I want most of the numbers to be close to
the middle, and only some at either high or low ends of the spectrum

eg if the range is 0-100, I want 70% of the numbers to be betwee 40 and
60.

From my vague memories from school maths, I kind of want to implement a
bell curve and use various standard deviation functionality. From my
description you can probably figire out I have fairly hazy recollections


Would someone be kind enough to direct me to a link that could give me
more information and/or some examples?

I don't know anything about nextGaussian like the other people suggested -
but if you could generate a curve of the probabilities of every location
that you want, and then integrate, you could solve for it and use a random
floating point number between 0 and 1. (This is fairly simply calculus.)