ODE Disease Dynamics Mini-Lab (Logistic growth / SI / SIR / SEIR)

This project is a teaching-oriented mini framework for deterministic epidemic models using deSolve. Goal: help understand model structure and parameter meanings, while supporting batch parameter sweeps that automatically output figures and summary tables.

Models

• Logistic growth (population dynamics)
• SI
• SIR
• SEIR

All transmission models use frequency-dependent transmission: beta * S * I / N.

A detailed interpretation of model equations for each model is available in: docs/model_equations.md

What you get

• Single-run trajectories for each model
• Parameter sweeps (grid runs) with:
  • faceted trajectory plots (one panel per scenario)
  • summary metrics tables (CSV): peak, peak time, attack rate, time-to-threshold, etc.

Quick start

Open R in the project root, then run:

install.packages("deSolve")

# Optional but recommended for stable PNG output on Windows
install.packages("ragg")

source("scripts/01_run_examples.R")

Outputs will be written to:

• outputs/figures/ To see more in example_outputs/.

• outputs/tables/

Example results are provided in example_outputs/. Running scripts/01_run_examples.R will generate new outputs in the outputs/ directory.

Parameter sweep strategies (less "random tuning", more interpretable)

To keep interpretation clean, each sweep fixes everything except the parameter of interest:

• Logistic: scan r
• SI: scan beta (and optional scan initial I0)
• SIR: fix gamma, scan beta so that R0 = beta/gamma changes transparently
• SEIR: fix beta and gamma, scan sigma (latent period = 1/sigma)

Note on "NN-like tuning": For mechanistic ODE models, we usually prefer principled sweeps (one-factor-at-a-time, or grid / Latin hypercube) plus interpretable outputs, rather than ad-hoc trial-and-error. And this ensures that the effect of each parameter (e.g., β, γ, σ) on epidemic dynamics can be interpreted unambiguously.

Results interpretation

A detailed interpretation of each model dynamics and parameter effects
is available in: docs/results_interpretation.md

This includes explanations of:

• peak prevalence and attack rate
• the role of (R_0)
• latent period effects in SEIR
• why t_50 and t_90 can be NA

Project layout

• R/ reusable functions (models, simulation, metrics, plotting)
• scripts/ one-click reproduction scripts
• outputs/ generated figures and tables

Ideas to extend

• Compare same R0 but different (beta, gamma) to show time-scale effects
• Add SIS, seasonal beta(t), interventions, or observation noise
• Export long-format data and use ggplot2 facets if you prefer