[ROOT-6666] Problem with 2-sample Anderson Darling test Created: 09/Sep/14  Updated: 30/Sep/14  Resolved: 16/Sep/14

Status: Closed
Project: ROOT
Component/s: None
Affects Version/s: None
Fix Version/s: 5.34/22

Type: Bug Priority: High
Reporter: Lorenzo Moneta Assignee: Lorenzo Moneta
Resolution: Fixed Votes: 0
Labels: None
Environment:

all


Attachments: File testAD_R.C     PDF File testAD_R.pdf    
Development:

 Description   

The 2 sample Ad test does not work and produces wrong results
See comparison with R using attached macro (using ROOT-R from
http://root.cern.ch/drupal/content/how-use-r-root-root-r-interface
The result is the p values obtained using 500 different pair of data samples obtained from the same normal distribution

Also the time increases exponentially for large sample sizes



 Comments   
Comment by Lorenzo Moneta [ 16/Sep/14 ]

This is now fixed in master and 5.34 patches. Some code has been taken from R package k-Samples
The timing is slightly better than previous implementation, but still going as N^2 when're N is the size of the samples.

Generated at Sat Sep 21 06:37:47 CEST 2019 using Jira 7.13.1#713001-sha1:5e06076c2d215a6f699b7e5c90ab2fae7ba5a1ce.