lgpr

lgpr performs nonparametric inference of covariate effects in longitudinal data using additive Gaussian processes to model nonlinear, temporal, and heterogeneous effects relevant to disease progression studies.

Key Features:

• Additive Gaussian Processes: Employs additive Gaussian processes for nonparametric modeling of covariate effects.
• Categorical and Continuous Covariates: Handles both categorical and continuous covariates within the Gaussian process framework.
• Nonlinear Interactions: Detects nonlinear interactions between covariates and captures both linear and nonlinear effect components.
• Handling of Temporal Uncertainty: Accounts for temporal uncertainty including rapid onset of effects near unobservable initiation times.
• Heterogeneous Covariate Effects: Models subject-specific heterogeneity in effect magnitudes across individuals.
• Flexible Observation Models: Implements observation models tailored to various types of biomedical data.
• Interpretability: Emphasizes interpretable effect inference to clarify contributions of covariates over time.

Scientific Applications:

• Longitudinal disease progression studies: Infer covariates that influence disease trajectories over time.
• Epidemiology: Model temporal and nonlinear risk factor effects in longitudinal population studies.
• Clinical trials: Analyze time-varying treatment and covariate effects in repeated-measures trial data.
• Personalized medicine: Quantify individual-level heterogeneity in covariate effects for personalized risk assessment.

Methodology:

Uses additive Gaussian processes to nonparametrically model linear and nonlinear covariate effects and interactions, incorporates observation models for different biomedical data types, accounts for temporal uncertainty near initiation times, and models heterogeneity of covariate effects across subjects.