| 1 |
# -*- org -*- |
| 2 |
#+CATEGORY: randist |
| 3 |
|
| 4 |
* ACM Computing Surveys (CSUR) |
| 5 |
Volume 39 , Issue 4 (2007) |
| 6 |
|
| 7 |
Gaussian random number generators |
| 8 |
David B. Thomas, Wayne Luk, Philip H.W. Leong, John D. Villasenor |
| 9 |
Article No. 11 |
| 10 |
|
| 11 |
|
| 12 |
* add Erlang dist back in? |
| 13 |
|
| 14 |
* DONE, for mu. |
| 15 |
|
| 16 |
Note that we need to get rid of mu when it is not the mean. |
| 17 |
|
| 18 |
From: Brian Gough <[email protected]> |
| 19 |
To: [email protected] |
| 20 |
Cc: [email protected] |
| 21 |
Subject: Re: Pareto Distribution |
| 22 |
Date: Sun, 9 Jul 2000 20:05:03 +0100 (BST) |
| 23 |
|
| 24 |
Yes, we should adopt the conventions from a standard reference book -- |
| 25 |
the existing functions are drawn from a variety of sources, mostly |
| 26 |
Devroye's book on Random Variates (which is public domain, but not |
| 27 |
available electronically unfortunately). Maybe the three volumes of |
| 28 |
Johnson & Kotz on Univariate Distributions would do, for |
| 29 |
example. Patches are welcome from anyone who wants sort this out. |
| 30 |
|
| 31 |
Keith Briggs writes: |
| 32 |
> Another thing to think about: some of the other distributions |
| 33 |
> have a argument `mu' to the C function which is a parameter |
| 34 |
> which is not the mean. This is non-standard and confusing. |
| 35 |
> (Also, in the Pareto function, `a' is normally called beta, |
| 36 |
> `b' is normally called alpha.) |
| 37 |
> |
| 38 |
> Keith |
| 39 |
> |
| 40 |
> +-------------------------------------------------------------------+ |
| 41 |
> | Dr. Keith M. Briggs, Complexity Research Group, BT Research Labs. | |
| 42 |
> | Adastral Park admin2 pp5, Martlesham Heath, IP5 3RE, Suffolk, UK | |
| 43 |
> | Tel. 01473 641 911 Fax. 01473 647 410. Home tel: 01473 625 972 | |
| 44 |
> | www.bt.com | personal homepage: www.labs.bt.com/people/briggsk2/ | |
| 45 |
> +-------------------------------------------------------------------+ |
| 46 |
|
| 47 |
|
| 48 |
* The exponential power distribution method could be speeded up by |
| 49 |
using a rational function approximation for the rejection scaling |
| 50 |
parameter. |
| 51 |
|
| 52 |
* Do something about the possibility of the user providing invalid |
| 53 |
parameters (e.g. negative variance etc). Not sure what to do though, |
| 54 |
since returning an error code is not possible. Maybe just return zero. |
| 55 |
We should return NAN in this case, and for the CDFs. |
| 56 |
|
| 57 |
* Add the triangular distribution. |
| 58 |
|
| 59 |
* Look at Marsaglia & Tsang, "The Monte Python Method for generating |
| 60 |
random variables", ACM TOMS Vol 24, No 3, p341 |
| 61 |
|
| 62 |
and the paper on the Ziggurat Method: Journal of Statistical Software, |
| 63 |
Volume 05 Issue 08. George Marsaglia and Wai Wan Tsang. "The ziggurat |
| 64 |
method for generating random variables" |
| 65 |
|
| 66 |
* Should 0 be included in distributions such as the exponential |
| 67 |
distribution? If we want a consistent behaviour, is it included in |
| 68 |
others? Note that 1-gsl_rng_uniform() can have a slight loss of |
| 69 |
precision when the random float is small. |