The relationship between confidence intervals and distributions of estimators for parameters of deterministic models

Konstadia Lika, Sebastiaan A.L.M. Kooijman*

*Corresponding author for this work

Research output: Contribution to JournalArticleAcademicpeer-review

Abstract

A Symmetric-Bounded (SB) method for parameter estimation enables estimating parameters in cases where maximum-likelihood (ML) methods are unsuitable. We here extend the SB-method to quantify the accuracy of point estimates and to link profile-based confidence intervals to the distribution of parameter estimators. We compare ML and SB methods for parameter estimates of Weibull and exponential models using Monte-Carlo-generated data. The SB-method performs at least as well as the ML-method. The SB-method is subsequently applied to real-world biological data (two coupled growth trajectories) where ML is unsuitable to quantify the accuracy of four underlying metabolic parameters of a deterministic growth model. The model fits the data perfectly. However, two of these parameters turn out to have narrow confidence intervals, while the remaining two do not. This discrepancy is elucidated by the shapes of the 2D confidence regions, which reveal the interdependence of the latter two parameters. Recommendations are proposed for increasing accuracy of parameters for mechanistic models for biological processes.

Original languageEnglish
Article number110645
Pages (from-to)1-11
Number of pages11
JournalEcological Modelling
Volume490
Early online date16 Feb 2024
DOIs
Publication statusPublished - Apr 2024

Bibliographical note

Publisher Copyright:
© 2024 The Authors

Funding

We would like to thank Dr Starrlight Augustine for her constructive feedback and suggestions. We also thank two anonymous reviewers for their comments and suggestions which helped to improve this manuscript.

Keywords

  • Add_my_Pet collection
  • Dynamic energy budget theory
  • Model plasticity
  • Practical identifiability
  • Profile-based interval estimates
  • Symmetric bounded loss function

Fingerprint

Dive into the research topics of 'The relationship between confidence intervals and distributions of estimators for parameters of deterministic models'. Together they form a unique fingerprint.

Cite this