Models

Inference

This directory contains all of the code necessary to estimate the parameters of each model, simulate the latent subtype-level infection dynamics, and perform model validation. Please refer to the Introductory page for a step-by-step overview of the manuscript analysis. Breifly, the worfklow for the model inference is as follows:

1. Fit the sub-model of the short-term titer dynamics Navigate to the Short_term_sub_model directory. Infer the parameters that govern the short-term titer boosting for children and adults with H1N1pdm09 and H3N2. Then, fix these parameters in the full longitudinal models (step 2).

2. Fit the full single-subtype models of the immune dynamics Navigate to the Single_strain_model directory. First, fit the transmission rate for each subtype to the full data. Then, infer the parameters that govern the dynamics of protection.

3. Fit the full multi-subtype model to infer the duration of heterosubtypic protection Navigate to the Mulit_strain_model directory. Fix the parameters that govern the single-strain dynamics of each subtype based on the output of steps 1 and 2. Then infer the rate of waning of heterosubtypic protection.

Simulations

The Simulations subdirectories contain the code to simulate the best-fit models using the inferred parameters and to produce the figures that appear in the text from the model simulations.

References

1. King AA, Nguyen D and Ionides EL (2015) Statistical inference for partially observed Markov processes via the R package pomp. arXiv preprint arXiv:1509.00503.

2. Ionides EL, Breto C, Park J, Smith RA, King AA (2017) Monte Carlo profile confidence intervals for dynamic systems. Journal of The Royal Society Interface 14(132).