Skip to content

Validation of R

2 messages · Brett Magill, Thomas Lumley

#
The national institute of standards and technology offers reference data sets and expected results for various statistical procedures using these data sets.  From the web site:

"The purpose of this project is to improve the accuracy of statistical software by providing reference datasets with certified computational results that enable the objective evaluation of statistical software."  

     http://www.itl.nist.gov/div898/strd/


-------Original Message-------
From: "Pikounis, Bill" <v_bill_pikounis at merck.com>
Sent: 04/17/03 07:53 AM
To: 'Rob Lambkin' <r.lambkin at retroscreen.com>, r-help at stat.math.ethz.ch
Subject: RE: [R] Validation of R
Notwithstanding the disclaimer automatically appended below by my "Big
Pharma" member IT dept on all my email sends, I have not had this problem,
or perhaps more accurately, not allowed it to be a problem when work I
have
done in R has made it into drug application filings and responses to FDA
and
European regulatory agencies.

"Validation" of software is an ill-defined concept, so I am afraid I
cannot
offer anything like a concrete "how-to", not would I be surprised if
anyone
else can.  What I would like to suggest is to (1) ask your vendor
companies
what specifically they are concerned about, (2) benchmark some guidelines
on
how you all or others have "validated" other software.

If you are looking for extensive documentation on whats/hows/whys of R, it
already has it.  If you are looking for it to compute the same values as
"validated" software within realistic numeric accuracy for your
procedures,
that is straightforward to do.  And the ultimate key is that anyone can
look
at the source code and have a high probability to get it to run on any
reasonably current system, and even many systems not so current. 

On a visible, continuous (daily), *OPEN* basis, there is ongoing review
and
input from the R user community, as well as all the highest standards of
software engineering that are met by the R core team and other developers. 
R
clearly stands up to rigorous, scholastic scrutiny. In my very grateful
view, this makes R at least as reliable as commercial vendor software that
claims "validation" or "compliance", etc., ...and probably, more reliable. 


Hope that helps.
Bill
----------------------------------------
Bill Pikounis, Ph.D.
Biometrics Research Department
Merck Research Laboratories
PO Box 2000, MailDrop RY84-16  
126 E. Lincoln Avenue
Rahway, New Jersey 07065-0900
USA

v_bill_pikounis at merck.com

Phone: 732 594 3913
Fax: 732 594 1565
______________________________________________
R-help at stat.math.ethz.ch mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-help
#
On Thu, 17 Apr 2003, Brett Magill wrote:

            
Yes, but the NIST tests I have seen (possibly an unrepresentative subset)
have been testing IMO the wrong thing. That is, they are good for finding
out where the rounding errors come in when the procedures are really
stressed.

They say
 "In response to industrial concerns about the numerical accuracy of
  computations from statistical software, the Statistical Engineering and
  Mathematical and Computational Sciences Divisions of NIST's Information
  Technology Laboratory are providing datasets with certified values for a
  variety of statistical methods."

In practice I think there's more danger from the wrong calculations being
done rather that from the results being accurate to 6 not  10
digits. Or from the wrong maximum being found in a multi-modal function,
which again is difficult to test.

It's perhaps also worth noting that the worst situation I know of in
recent years arose from accepting a poor default for a user-adjustable
precision setting (in S-PLUS, not that it really matters where).


	-thomas