Hello,I found some disturbing results for the CORELATION COEFFICIENT. See the attached Test Cases, where a corelation coeff. of -1.64 was calculated. This is definitely wrong.
I also detected, that the mean is calculated sometimes NOT as accurate as it could. IF the calculation algorithm would first sort the values (as absolute values), the result would be even more exact. See the 2nd Test Case.
I detailed some of these reasons on the OpenOffice Spreadsheet mailing list, see also my posts there (http://sc.openoffice.org/servlets/SummarizeList?listName=dev), especially in the October mailing list (http://sc.openoffice.org/servlets/BrowseList?listName=dev&from=2006-10-01&to=2006-10-31&count=42&by=date&first=21&windowSize=20&selectedPage=2 , thread on "Statistical Functions Implementations").
This thread is continued in November, too. (http://sc.openoffice.org/servlets/BrowseList?listName=dev&by=date&from=2006-11-01&to=2006-11-30&first=1&count=6)The National Institute of Standards has some Test Cases, too. See my post in November for further details.
Kind regards, Leonard Mada
Attachment:
Calc-Test-Sheet-CORREL.ods
Description: application/vnd.oasis.opendocument.spreadsheet
Attachment:
Calc-Test-Sheet-SUM.ods
Description: application/vnd.oasis.opendocument.spreadsheet