Premium
WAM (Wald's Approximation Method): a user‐friendly software program for efficient covariate model building
Author(s) -
Kowalski K. G.,
Wang W.,
Hermann D.
Publication year - 2004
Publication title -
clinical pharmacology and therapeutics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.941
H-Index - 188
eISSN - 1532-6535
pISSN - 0009-9236
DOI - 10.1016/j.clpt.2003.11.158
Subject(s) - covariate , nonmem , subroutine , computer science , population , software , file transfer protocol , statistics , algorithm , mathematical optimization , mathematics , programming language , machine learning , operating system , demography , the internet , sociology
A new software utility (WAM) based on the efficient covariate model building algorithm proposed by Kowalski and Hutmacher (2001) is presented and will be made available on the NONMEM repository ( ftp://ftp.globomaxnm.com/Public/nonmem/ ). The WAM software is a user‐friendly freeware utility that ranks all 2 k possible covariate sub‐models (presence or absence of k covariate parameters) based on Wald's approximation to Schwarz's Bayesian Criterion (SBC). The utility generates NONMEM runs for the top M parsimonious models (i.e., highest M ranked SBCs where M is typically set to 10 or 15) and compares the actual (NONMEM) and approximate (Wald) SBCs to verify the adequacy of the approximation. The final parsimonious model is chosen corresponding to the model with the highest actual SBC among the conditionally ranked top M models. The utility requires the following: Minor coding conventions in the NONMEM control stream for the full model (all k covariate parameters included in the model simultaneously), Successful COV step estimation for the full model run, and Use of an external subroutine ($SUBROUTINE command) and/or verbatim code to output the estimates and covariance matrix from the full model fit.The WAM utility was written in C and was developed, tested and validated in a Windows PC environment; however, it is expected that the utility will be portable and should run under any operating system where NONMEM is supported. Clinical Pharmacology & Therapeutics (2004) 75 , P42–P42; doi: 10.1016/j.clpt.2003.11.158