RE: [NMusers] NONMEM6: BQL and YLO

Dear Pavel,
 
Your are apparently in luxorious position that you can lay your hand on
the actual values in case of BQL values. Modeling using all data
regardless of censoring is than possible, by inclusion of an additive
error; this preserves more of the information contained in BQL samples.
The resulting additive error might be viewed upon in terms of a
detection limit (actually the squared standard deviation thereof),
assuming an otherwise perfect model of course.
 
Best regards,
Jeroen

        -----Original Message-----
        From: owner-nmusers_at_globomaxnm.com
[mailto:owner-nmusers_at_globomaxnm.com] On Behalf Of Ludden, Thomas (MYD)
        Sent: Monday, 08 January, 2007 15:30
        To: nonmem_at_optonline.net
        Cc: nmusers_at_globomaxnm.com
        Subject: RE: [NMusers] NONMEM6: BQL and YLO
        
        
        Dear Pavel and NONMEM6 users,
         
        I believe the example referred to implements Stuart Beal's
strategy II (J. Pharmacokin. Pharmacodyn. 32:213-243,2005) described
specifically on page 230-231 of that article as well as in a previous
article (J. Pharmacokin. Pharmacodyn. 28:481-504, 2001). That approach
involves deletion of the DV<LOQ from the calculation of the likelihood
(MDV=1) and corrects the likelihood for all remaining observations in
the data set. Examine the PR_Y values for observations that are above
but near the LOQ. This is at least a somewhat better estimate of the
likelihood for the data remaining in the data set. For the users
information, Stuart allows the calculation of the probabilities
associated with DV<LOQ, MDV=1 if measured values are availble but, as
you note, these do not influence -2LL. In most data sets, DV<LOQ values
are not reported quantitatively.
         
        Hope this is helpful.
         
        Tom

  _____

        From: owner-nmusers_at_globomaxnm.com
[mailto:owner-nmusers_at_globomaxnm.com] On Behalf Of nonmem_at_optonline.net
        Sent: Friday, January 05, 2007 9:12 PM
        To: nmusers_at_globomaxnm.com
        Subject: [NMusers] NONMEM6: BQL and YLO
        
        

        Hello NONMEM6 users,

        At one conference new features of NONMEM6 were presented. It
was stated that NONMEM6 takes care of BQL values and improves
likelihood. I tried to find information in the manuals and found the
YLO example and the following statement: "Last observation record
(below) has DV < LOQ (LOQ = 1) and MDV = 1. This observation does =
not
contribute to the -2LL estimate. However, probability that DV is > LOQ
can be tabled using PRB=PR_Y." Therefore (?), taking care of BQLs in
the YLO examlpe does not improve the results.

        1. Is there a way to utilize BQLs and improve likelihood?

        2. How helpful is the estimate of PR_Y?

        Thank you!

        Pavel

        
        
        
        
________________________________________________________________________
_______________________________________________________________
        
          
        
         
        This e-mail transmission may contain confidential or legally
privileged information that is intended only for the individual or
entity named in the e-mail address. If you
        are not the intended recipient, you are hereby notified that any
disclosure, copying, distribution, or reliance upon the contents of this
e-mail is strictly prohibited.
        If you have received this e-mail transmission in error, please
reply to the sender, so that ICON can arrange for proper delivery, and
then please
        delete the message. Thank You.

        




         









        


This message, including attachments, is confidential and may be =
privileged. If you are not an intended recipient, please notify the =
sender then delete and destroy the original message and all copies. You =
should not copy, forward and/or disclose this message, in whole or in =
part, without permission of the sender.

Received on Tue Jan 09 2007 - 17:05:14 EST