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

To analyse if the very recent decay of infection in Germany is significant, we used the three β-values as the outcome variable in three different appropriate GAMMs. In particular, we used a 2D thin spline regression depending on geographical coordinates to reduce spatial autocorrelation [18] , and introduced the federal state as a random intercept. In particular, for β 1 and β 3 we used the Gaussian probability distribution, whereas for β 2 we applied a Beta-distribution since this value is restricted between zero and one [3, 9] . Furthermore, for β 1 and β 2 , data from different ADs have been weighted (using a priori regression weights) with respect to the total number of cases in the corresponding AD, whereas for β 3 the inverse variance of the estimated regression coefficient has been used as a weight ('inverse variance weighting' [6] ). Furthermore, the federal state has been used as a random intercept in all three models.