commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Phil Steitz" <p...@steitz.com>
Subject RE: [math] Questions regarding probability distributions
Date Thu, 14 Oct 2004 00:00:27 GMT
I agree with all points below.  It occured to me later this AM that exposing the right parameters
meant that the implementations needed to be separate.  I now agree with you and Mark that
this would be too tricky to implement correctly while presenting the right interface to the
user.
 
What is you opinion on the DiscreteDistribution int vs. double issue?
 
Phil
 
-----Original Message----- 
From: Brent Worden [mailto:brent@worden.org] 
Sent: Wed 10/13/2004 2:08 PM 
To: Jakarta Commons Developers List 
Cc: 
Subject: Re: [math] Questions regarding probability distributions



	Here's my basis for the exponential, chi-squared, and gamma distribution design (forgive
me for reiterating some points made by others):
	
	Exponential is separate because both the CDF and inverse CDF have closed form computations.
 This provides better performance and higher accuracy than the iterative method approximation.
	
	Both exponential and chi-squared are separate from gamma to provide access only to the parameters
specific to each distribution.
	
	In chi-squared case, extension could have easily been use but this would have exposed the
alpha and beta properties in chi-squared.  This would allow users to modify the properties
in an uncontrolled fashion, possibly resulting in a disfunctional distribution.  Thus, encapsulation
was used to hide the gamma properties and only expose the relavent chi-squared properties.
	
	
	---------------------------------------------------------------------
	To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
	For additional commands, e-mail: commons-dev-help@jakarta.apache.org
	
	

Mime
View raw message