Comparing COVID‐19 incidences longitudinally per economic sector against the background of preventive measures and vaccination

In the COVID‐19 pandemic, workplace transmission plays an important role. For this type of transmission, the longitudinal 14‐day incidence curve of SARS‐CoV‐2 infections per economic sector is a proxy. In Belgium, a census of confirmed 14‐day incidences per NACE‐BEL sector level three is available from September 2020 until June 2021, encompassing two waves of infections. However, these high‐dimensional data, with a relatively small number of NACE‐BEL sectors, are challenging to analyze. We propose a nonlinear Gaussian–Gaussian model that combines parametric and semi‐parametric elements to describe the incidence curves with a small set of meaningful parameters. These parameters are further analyzed with conventional statistical methods, such as CCA and linear models, to provide insight into predictive characteristics of the first wave for the second wave. Those nonlinear models classify economic sectors into three groups: sectors with two regular waves of infections, sectors with only a first wave and sectors with a more irregular profile, which may indicate a clear effect of COVID‐19 vaccination. The Gaussian–Gaussian model thus allows for analyzing and comparing incidence curves and to bring out key characteristics of such curves. Finally, we consider in which other settings the proposed approach could be applied, together with possible pitfalls.