dynr.mi

dynr.mi performs multiple imputation for dynamic systems models using Multivariate Imputation by Chained Equations (MICE) to handle missingness in intensive longitudinal data and enables estimation and visualization of linear and nonlinear discrete- and continuous-time models.

Key Features:

• MICE integration: Leverages Multivariate Imputation by Chained Equations (MICE) to create multiple imputations for missing data in intensive longitudinal data (ILD).
• Dynamic systems support: Fits and visualizes both linear and nonlinear dynamic systems models in discrete and continuous time.
• Non-ignorable missingness handling: Allows specification of models with potentially non-ignorable missingness in dependent variables and covariates.
• Multiple imputed datasets: Generates multiple imputed datasets and pools estimation results across them to produce final inferential estimates.
• Convergence diagnostics: Provides convergence diagnostic checks to determine the appropriate number of imputation iterations.
• Vector autoregressive models: Supports estimation of vector autoregressive models for multivariate time series within dynamic modeling.

Scientific Applications:

• Intensive longitudinal data analysis: Address missing data and estimate dynamic relationships in ILD studies.
• Ambulatory physiology and affect: Model the interplay between ambulatory physiological measures and self-reported affect (valence and arousal) using vector autoregressive frameworks.
• Missing-data sensitivity: Compare inferential outcomes obtained via multiple imputation versus listwise deletion of incomplete records.
• Inference from pooled estimates: Derive final parameter estimates and inference by pooling results across multiple imputations.

Methodology:

Creates multiple imputed datasets using MICE, fits linear and nonlinear dynamic systems (including vector autoregressive models) in discrete and continuous time, performs convergence diagnostic checks for imputation iterations, and pools estimation results across imputations to obtain final inferential estimates.