Campus Access Only

All rights reserved. This publication is intended for use solely by faculty, students, and staff of University of the Pacific. No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, now known or later developed, including but not limited to photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author or the publisher.


Evaluation of 10-fold cross validation and prediction error sums of squares statistic for population pharmacokinetic model validation

Date of Award


Document Type

Thesis - Pacific Access Restricted

First Advisor

Paul Williams

First Committee Member

James Uchizono

Second Committee Member

Marcus Ravnan


It was the objective of the current study to evaluate the ability of 10-fold crossvalidation and prediction error sum of squares (PRESS) statistic to identify population pharmacokinetic models (PPKM) that were estimated from data without influence observations versus PPKMs from data containing influence observations. The evaluation of 10-fold cross validation and PRESS statistic from Leave-one-out cross-validation for PPK model validation was performed in 3 Phases. In Phase 1 model parameters (theta and clearance) were estimated for datasets with and without influence observations. It was found that influence observations caused an over-estimation of the model parameters.

In Phase II the statistics from 10-fold and leave-one-out cross validation methods were used to detect models developed from influence data. The metrics of choice are RATIOK and RATIOPR statistics that can be used to identify models developed from influence data and these metrics may then find applicability across differing drugs and models. A cut-off value of 1.05 for RATIOK and RATIOPR was proposed as a discrete breakpoint to classify models that were generated from influence data versus noninfluence data.

In Phase III data analysis was carried out using logistic regression and the sensitivity and specificity of Leave-one-out and 10-fold cross-validation methods were evaluated. It was found that RATIOK and RATIOPR were significant predictors when used individually in the model. Multicollinearity was detected when RATIOK and RATIOPR were present in the model at the same time. In terms of sensitivity and specificity both 10-fold cross validation and leave-one-out cross validation showed similar performance.



This document is currently not available here.

To access this thesis/dissertation you must have a valid email address and log-in to Scholarly Commons.

Find in PacificSearch



If you are the author and would like to grant permission to make your work openly accessible, please email