What influences COVID-19 infection rates: A statistical approach to identify promising factors applied to infection data from Germany

Moritz Mercker; Uwe Betzin; Dennis Wilken

Overall, there is a very coherent picture of the spatial distribution of the different β-values: In all cases, the distribution is spatially inhomogeneous, showing in particular in the centre of Germany (on the boarder of Türinghen, and Hessen) a lower decay in infection rates compared to the average development. Indeed, both states show distinct larger beta values also in our analysis on the state-level (3). Only the cross-validation based GAMM-plot for β 3 (4 third row, plot on the right-hand side) shows a much coarser spatial resolution. A possible explanation is that for the calculation of β 3 much less data (namely infection data from only the last two weeks) have been used, which may lead to a larger unexplained variance, which finally leads to a more conservative estimate of spatial heterogeneity during the cross-validation procedure.