Abstract:Non-linear mixed effect models (NLMEMs) are widely used for the analysis of longitudinal data. To design these studies, optimal design based on the expected Fisher information matrix (FIM) can be used instead of performing time-consuming clinical trial simulations. In recent years, estimation algorithms for NLMEMs have transitioned from linearization toward more exact higher-order methods. Optimal design, on the other hand, has mainly relied on first-order (FO) linearization to calculate the FIM. Although efficient in general, FO cannot be applied to complex non-linear models and with difficulty in studies with discrete data. We propose an approach to evaluate the expected FIM in NLMEMs for both discrete and continuous outcomes. We used Markov Chain Monte Carlo (MCMC) to integrate the derivatives of the log-likelihood over the random effects, and Monte Carlo to evaluate its expectation w.r.t. the observations. Our method was implemented in R using Stan, which efficiently draws MCMC samples and calculates partial derivatives of the log-likelihood. Evaluated on several examples, our approach showed good performance with relative standard errors (RSEs) close to those obtained by simulations. We studied the influence of the number of MC and MCMC samples and computed the uncertainty of the FIM evaluation. We also compared our approach to Adaptive Gaussian Quadrature, Laplace approximation, and FO. Our method is available in R-package MIXFIM and can be used to evaluate the FIM, its determinant with confidence intervals (CIs), and RSEs with CIs.

Influence Analysis of Additive Mixed-Effects Nonlinear Regression Models Via EM Algorithm

Mixed Model Approaches for Detecting Influential Observations in Genetic Data Analysis

Influence Analysis of Generalized Partially Linear Mixed Models

Influence Analysis on Semiparametric Generalized Linear Mixed Model

Additive mixed models with approximate Dirichlet process mixtures: the EM approach

Influence Diagnostics in Meta-Regression Model

Local Influence for Semiparametric Generalized Linear Model

Machine learning embedded EM algorithms for semiparametric mixture regression models

Em Algorithm of the Truncated Multinormal Distribution with Linear Restriction on the Variables

Simulation Study on Local Influence Diagnosis for Poisson Mixed-Effect Linear Model

Estimation and Influence Diagnostics for the Multivariate Linear Regression Models with Skew Scale Mixtures of Normal Distributions

Estimating mixed-effects state-space models via particle filters and the EM algorithm

Statistical Machine Learning EM algorithm Lecture Notes 13 : EM algorithm

The Influence Function of Semiparametric Estimators

Regularized EM Algorithms: A Unified Framework and Statistical Guarantees

The use of the EM algorithm for regularization problems in high-dimensional linear mixed-effects models

Empirical Likelihood Methods Based on Influence Functions

An MCMC method for the evaluation of the Fisher information matrix for non-linear mixed effect models

Benefits of over-parameterization with EM

Computational Aspects Of Analyzing Random Effects Longitudinal Models

Influence diagnostics in Birnbaum-Saunders nonlinear regression models