Skip Menu |

This queue is for tickets about the Statistics-RankCorrelation CPAN distribution.

Report information
The Basics
Id: 7752
Status: resolved
Worked: 4 min
Priority: 0/
Queue: Statistics-RankCorrelation
People
Owner: gene [...] cpan.org
Requestors: Vl_Babenko [...] softhome.net
Cc:
AdminCc:
Bug Information
Severity: (no value)
Broken in: (no value)
Fixed in: (no value)

History Show all quoted text
  Thu Sep 23 12:15:24 2004 Guest - Ticket created
Subject: simplified formula for spearman
Greetings; I tried to compare Spearman rank correlation in case of tied ranks with some other programs. I observed the discrepancy, rather significant in some cases, up to 0.5. The discrepancy arises when the ranks are tied. Regards, Vladimir Babenko
  Thu Sep 23 18:02:46 2004 gene [...] cpan.org - Given to GENE
  Thu Sep 23 18:02:46 2004 gene [...] cpan.org - Status changed from 'new' to 'open'
  Sat Mar 24 07:00:45 2007 gene [...] cpan.org - Subject changed from 'simplified formula for spearman' to 'discrepancy seen (in Matlab) with tied ranks'
  Sat Mar 24 07:04:10 2007 gene [...] cpan.org - Reference by ticket #4237 added
  Sat Mar 24 14:00:04 2007 gene [...] cpan.org - Correspondence added
On Thu Sep 23 12:15:24 2004, guest wrote: Show quoted text
> I tried to compare Spearman rank correlation in case of tied ranks with some other
programs. I observed the discrepancy, rather significant in some cases, up to 0.5. The discrepancy arises when the ranks are tied. Show quoted text
> Regards, > Vladimir Babenko
What are examples of the data vectors which produce these discrepencies? Gene
  Sun Dec 02 21:36:05 2007 gene [...] cpan.org - Correspondence added 4 min
RT-Send-CC: Vl_Babenko [...] softhome.net
After a bit of research, I discovered that computation with "tied ranks" fails. I've uploaded a _dev version that illustrates this with failing tests and a warning in the POD. So far I cannot figure out the answer...
  Mon Dec 03 02:34:41 2007 gene [...] cpan.org - Correspondence added
RT-Send-CC: Jon.Schutz [...] youramigo.com
I added the working code given to me by Jon Schutz, which fixed all but one of the failing pearson() tests. Voila! Thanks Jon. :) Gene Boggs Software Engineer at-large
  Mon Dec 03 02:36:45 2007 gene [...] cpan.org - Status changed from 'open' to 'resolved'