Documentation of the gamma and lambda functionality in #Error

What would you like Pmetrics to do in the future?
Post Reply
tbouillon
Posts: 16
Joined: Sun Feb 22, 2015 11:51 pm
Location: Switzerland

Documentation of the gamma and lambda functionality in #Error

Post by tbouillon » Mon Nov 07, 2016 5:36 am

Geachte Collegas
I am abusing Lambda and Gamma to estimate additive or proportional errors (fixing the respective coefficients in the error term to 0 and 1).
The old LAPK technical reports and the literature in general highlight, that it is impossible "to separate the various sources of variability" with nonparametric methods. But that is effectively what I do and the program runs with meaningful estimates.
Questions:
a) is that what I do defendable?
b) if yes, how well is your gamma, lambda estimation documented, it should be a publication in its own right, apparently fixing one severe shortcoming of the nonparametric approach.
c) will I run into problems later, i.e. are the obj. fct. values reliable (probably yes) and how is gamma/lambda handled in the simulation based diagnostics?
Any help on this would be greatly appreciated, I need to know if I want to publish with Pmetrics.
Met vriendelijke groeten
Thomas

User avatar
mneely
Site Admin
Posts: 209
Joined: Tue Nov 08, 2011 11:08 am
Contact:

Re: Documentation of the gamma and lambda functionality in #Error

Post by mneely » Mon Nov 07, 2016 1:47 pm

I will get the other lab members to weigh in on a combined response, Thomas. Hold tight...

tbouillon
Posts: 16
Joined: Sun Feb 22, 2015 11:51 pm
Location: Switzerland

Re: Documentation of the gamma and lambda functionality in #Error

Post by tbouillon » Sun Dec 04, 2016 3:25 am

Dear Michael
Any news? Please help.
Quick motivation from 2009 (Savic, Karlsson):
The AAPS Journal, Vol. 11, No. 3, September 2009 (# 2009)
DOI: 10.1208/s12248-009-9138-8
Evaluation of an Extended Grid Method for Estimation Using Nonparametric
Distributions

p 623
One of the characteristics of majority of nonparametric
methods is that these methods are not able to estimate
simultaneously interindividual and residual variability. An
exception to this is NPAG implemented in USC*PACK
(3,4), for which applications (15,16) have reported estimated
residual error. The underlying methodology for this
estimation has to our knowledge not been published.
Therefore, it is a common procedure to assess the residual
variability a priori via parametric analysis, and subsequently
the nonparametric estimation is conditioned on this estimate.
Best regards
Thomas

Post Reply