area bug(?)

View: New views
3 Messages — Rating Filter:   Alert me  

area bug(?)

by Tom Gettys :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

Some parts of this message have been removed. Learn more about Nabble's security policy.

f(x)=dnorm(x,166.3, 26/sqrt(40))

 

area under path:

from: 173.06

to: 1000

area: 0.05 (correct)

 

when I add one more zero to the to: value, the area becomes 0.0826, which is incorrect

 

Tom Gettys

2929 Warren Street

Eugene, Oregon 97405

(541) 683-1940

 

To dream of the person you would like to be

is to waste the person you are.

 

 


------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables
unlimited royalty-free distribution of the report engine
for externally facing server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Graph-support mailing list
Graph-support@...
https://lists.sourceforge.net/lists/listinfo/graph-support

Re: area bug(?)

by Ivan Johansen :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

Tom Gettys wrote:
> f(x)=dnorm(x,166.3, 26/sqrt(40))
> area under path:
> from: 173.06
> to: 1000
> area: 0.05 (correct)
>
> when I add one more zero to the to: value, the area becomes 0.0826,
> which is incorrect

This is an unfortunately side effect of the algorithm (Simpson's
formula) used to calculate the area. When the range gets large rounding
errors starts to accumulate. When I find the time I plan to search for a
better algorithm.

Best regards
Ivan Johansen


------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables
unlimited royalty-free distribution of the report engine
for externally facing server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Graph-support mailing list
Graph-support@...
https://lists.sourceforge.net/lists/listinfo/graph-support

Re: area bug(?)

by Tom Gettys :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

No problem - I thought it might be something like that, but wanted to report
it in case it was a real bug.

Tom Gettys
2929 Warren Street
Eugene, Oregon 97405
(541) 683-1940
 
To dream of the person you would like to be
is to waste the person you are.
 
-----Original Message-----
From: Ivan Johansen [mailto:graph@...]
Sent: Monday, May 18, 2009 3:01 PM
To: graph-support@...
Subject: Re: [Graph-support] area bug(?)

Tom Gettys wrote:
> f(x)=dnorm(x,166.3, 26/sqrt(40))
> area under path:
> from: 173.06
> to: 1000
> area: 0.05 (correct)
>
> when I add one more zero to the to: value, the area becomes 0.0826,
> which is incorrect

This is an unfortunately side effect of the algorithm (Simpson's
formula) used to calculate the area. When the range gets large rounding
errors starts to accumulate. When I find the time I plan to search for a
better algorithm.

Best regards
Ivan Johansen


----------------------------------------------------------------------------
--
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables
unlimited royalty-free distribution of the report engine
for externally facing server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Graph-support mailing list
Graph-support@...
https://lists.sourceforge.net/lists/listinfo/graph-support


------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables
unlimited royalty-free distribution of the report engine
for externally facing server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Graph-support mailing list
Graph-support@...
https://lists.sourceforge.net/lists/listinfo/graph-support