Hello, I would extract some numbers according to a given probability distribution (at least, Normal, Exponential, Poisson and Bernoulli). Do you know if a similar library already exists in Ruby? I only found pseudo-random number generator. TIA Libra

on 2007-04-12 23:56

on 2007-04-13 00:35

If you don't find any library, you can always write your own using the technique outlined in section 1.3 of this paper. http://eeyore.ucdavis.edu/stat141/RNG.pdf Basically given the CDF F of some distribution (say, Poisson), you can generate random numbers from the Poisson distribution by doing: u = rand # uniformely generated random number y = Finv(u) -Szymon

on 2007-04-13 02:36

```
On 13 Apr 2007, at 06:55, Libra wrote:
> Libra
If you are comfortable with R, you can use RSRuby (http://
web.kuicr.kyoto-u.ac.jp/~alexg/rsruby/) to generate numbers from any
its distribution functions. For example, extracting 10 numbers from a
normal distribution with mean 0:
[alexg@powerbook]/Users/alexg(1): irb -rubygems
irb(main):001:0> require 'rsruby'
=> true
irb(main):002:0> r = RSRuby.instance
=> #<RSRuby:0x1c1510>
irb(main):003:0> r.rnorm(10)
=> [-0.418488870801272, 3.8442920287686, -2.23915973195997,
-1.1852964455658, 0.292187794718865, -0.0538355720349994,
-0.682675725085902, -0.179593328583021, 0.617238096715829,
-0.353627801112474]
The default R stats package has the following distributions built in
and you can use any R library to add more (I couldn't see Bernoulli
in this list, but maybe it has an alternative name?):
Beta(stats) The Beta Distribution
Binomial(stats) The Binomial Distribution
Cauchy(stats) The Cauchy Distribution
Chisquare(stats) The (non-central) Chi-Squared Distribution
Exponential(stats) The Exponential Distribution
FDist(stats) The F Distribution
GammaDist(stats) The Gamma Distribution
Geometric(stats) The Geometric Distribution
Hypergeometric(stats) The Hypergeometric Distribution
Logistic(stats) The Logistic Distribution
Lognormal(stats) The Log Normal Distribution
Multinomial(stats) The Multinomial Distribution
NegBinomial(stats) The Negative Binomial Distribution
Normal(stats) The Normal Distribution
Poisson(stats) The Poisson Distribution
SignRank(stats) Distribution of the Wilcoxon Signed Rank
Statistic
TDist(stats) The Student t Distribution
Tukey(stats) The Studentized Range Distribution
Uniform(stats) The Uniform Distribution
Weibull(stats) The Weibull Distribution
Wilcoxon(stats) Distribution of the Wilcoxon Rank Sum Statistic
ecdf(stats) Empirical Cumulative Distribution Function
survreg.distributions(survival)
Parametric Survival Distributions
Alex Gutteridge
Bioinformatics Center
Kyoto University
```

on 2007-04-13 02:50

On 13 Apr 2007, at 09:35, Alex Gutteridge wrote: >> TIA >> >> Libra > > If you are comfortable with R, you can use RSRuby (http:// > web.kuicr.kyoto-u.ac.jp/~alexg/rsruby/) to generate numbers from > any its distribution functions. A quick bit of googling found the missing Bernoulli distribution function: http://rss.acs.unt.edu/Rdoc/library/Rlab/html/Bernoulli.html It seems to be part of a CRAN package called Rlab: http://cran.r-project.org/src/contrib/Descriptions/Rlab.html You should be able to install and then access this library through RSRuby to get your complete set of functions. Alex Gutteridge Bioinformatics Center Kyoto University

on 2007-04-15 14:16

Libra wrote: > Libra R is great! maybe too great for your requirements rb-gsl is easy for example this will print you a nice gauss-shape graph from rb-gsl-1.8.3/samples/histogram/gauss.rb > > sigma, mean, height, = h.fit_gaussian > > x = GSL::Vector.linspace(-MAX, MAX, 100) > y = height*Ran::gaussian_pdf(x-mean, sigma) > GSL::graph(h, [x, y], "-T X -C -g 3") NICE! cauchy exponential power poisson etc. also available ruby - making maths easy

on 2007-08-23 23:55

From wikipedia : algorithm for generating pseudo poissons: def poisson(lambda) # init l = Math.exp(-lambda) k = 0 p = 1 while p >= l k += 1 u = rand p = p *u end return k -1 end only returns ints, but hey, they center around lambda gl. -Roger maw wrote: > Libra wrote: >> Libra > R is great! maybe too great for your requirements > rb-gsl is easy > > for example this will print you a nice gauss-shape graph > from rb-gsl-1.8.3/samples/histogram/gauss.rb >> >> sigma, mean, height, = h.fit_gaussian >> >> x = GSL::Vector.linspace(-MAX, MAX, 100) >> y = height*Ran::gaussian_pdf(x-mean, sigma) >> GSL::graph(h, [x, y], "-T X -C -g 3") > > NICE! > > cauchy exponential power poisson etc. also available > ruby - making maths easy

on 2007-08-25 06:06

Roger Pack wrote: > u = rand >> Libra wrote: >>> GSL::graph(h, [x, y], "-T X -C -g 3") >> NICE! >> >> cauchy exponential power poisson etc. also available >> ruby - making maths easy > There is also RSRuby, a way of calling the R libraries, including their probability distribution routines, from Ruby. I don't know anything about the ones in GSL, but I know the algorithms in R are first-rate, and I wouldn't code my own no matter *how* bored I got. :)