[Rd] header containing (PR#9171)

JLewis at wyeth.com JLewis at wyeth.com
Mon Oct 9 23:26:29 CEST 2006


Another example where the "Warning message: full precision was not
achieved in 'pnt'" is given is
  qt(.95,15,12)
which returns the value 17.631800001788577 in R2.4.0.  By symmetry,
this value should be identical to
  -qt(.05,15,-12)
which returns without warning the value 17.6318000017885623 in R2.4.0. 
To full DP precision, I believe that qt(.95,15,12) and -qt(.05,15,-12)
should be 17.631800001755318 (verified using the Maple
CDF(NonCentralStudentT()) function evaluated to 50 figures).
 
Since the difference between qt(.95,15,12) and -qt(.05,15,-12) is
orders of magnitude smaller than their differences from the correct
value, it seems that the decision on whether to give this warning is
being made based on criteria that the algorithm is not capable of
supporting.
 
Jerry W. Lewis
Wyeth BioPharma
Andover, MA
 
 

	[[alternative HTML version deleted]]




More information about the R-devel mailing list