Exposure-Lag-Response Modeling for Plant Disease Epidemiology

Supports modeling and interpretation of environmental exposure-lag-response relationships in plant disease epidemiology using distributed lag non-linear models (DLNMs). Provides functions for defining complete exposure profiles, evaluating model identifiability, summarizing lag-specific and period-specific effects, performing sensitivity analyses, simulating environmental exposure scenarios, predicting plant disease outcomes, quantifying yield and economic losses, and evaluating individual and ensemble models. Supports harmonized frequentist and Bayesian modeling workflows. Builds on the framework implemented in the 'dlnm' package and described by Gasparrini (2011) .

EpiExposure

EpiExposure is an R package for modeling and interpreting epidemiological exposure–lag–response relationships, simulating exposure profiles, and predicting plant disease outcomes from environmental drivers.

Core workflow

Why EpiExposure?

In plant disease epidemiology, environmental effects are often non-linear, cumulative, and time-dependent. Traditional approaches commonly summarize environmental conditions within predefined temporal windows, which may obscure delayed or continuously varying associations. EpiExposure provides a unified DLNM-based workflow for preserving exposure profiles, modeling lagged associations, and translating fitted models into epidemiologically interpretable summaries, simulations, and predictions.

What does it do?

EpiExposure enables you to:

• Model non-linear exposure–lag–response relationships
• Quantify lag- and period-specific environmental associations
• Define biologically meaningful epidemiological periods
• Summarise complete exposure profiles using the Exposure Cumulative Impact (ECI)
• Decompose cumulative exposure impacts into exact lag-specific contributions
• Simulate and compare hypothetical exposure–lag scenarios
• Predict expected disease outcomes from complete exposure profiles
• Compare and ensemble alternative DLNM specifications using cross-validation

Learning path

1. Define exposure–lag structures with define_exposures().
2. Build the DLNM design with build_design().
3. Fit the epidemiological model with fit_epidlnm().
4. Interpret exposure–lag relationships using summarise_effects().
5. Explore hypothetical exposure profiles with simulate_exposures().
6. Predict expected outcomes using predict_outcomes().
7. Evaluate alternative model structures with find_bestfit() and ensemble_bestfit().

Installation

EpiExposure is currently available from GitHub. Install the package with:

install.packages("pak")
pak::pkg_install("tomazrg/EpiExposure")

Then load the package:

library(EpiExposure)