Inference for Quasi-reaction Models with Covariate-Dependent Rates

Statistical models of quasi-reaction systems are typically described by constant reaction rates. This assumption is too restrictive in many applications, as rates may vary dynamically, spatially or within groups of the population. In this paper, we capture this heterogeneity with the inclusion of covariates in the dynamic model. In particular, we propose an extension of a recently developed latent event history model, by allowing log-reaction rates to be linearly dependent on a vector of covariates. We describe an inferential approach for parameter estimation of the resulting model and evaluate its performance via a simulation study. Finally, we show an illustration on COVID-19 data, where the approach is able to measure the effect of environmental factors and governmental interventions on the disease spreading and severity.