Fig. 3
Multivariable regression analysis of variables associated with COVID-19 cases. a results from all-subsets regression analysis to identify the best sub-model with a smaller list of variables, based on minimization of the AIC. Selected variables are highlighted in red (in addition to the intercept). The x-axis refers to the model size (number of variables in each sub-model), and y-axis lists all the variables tested as follows: Temp_Jan(oC,2016), temperatures in January in degrees Celsius in 2016; Temp_Feb(oC, 2016), temperatures in February in 2016; Temp_Mar(oC, 2016), temperatures in March, 2016; Temp_Apr(oC, 2016), temperatures in April, 2016; Urban%(2018), percentage of urban living population in 2018; Emp_service_%total(2018), percentage of total male employment in service sector in 2018; Emp_agri_%total(2018), percentage of total male employment in agriculture in 2018; Emp_ind_%total(2018), percentage of total male employment in industry in 2018; log-COVID_duration(May2020); duration (in days) between May 11, 2020 and the first reported COVID-19 case in a country (log10 scale); log_popdens(2018), population density in 2018 (log10 scale); log_Rain_Feb(mm,2018), rain in millimetres in February 2018 (log10 scale); Age_15_64yrs(2018), population between the ages 15 to 64 as percentage of the total population in 2018; >65yrs_%total(2018), population aged 65 and above as percentage of the total population in 2018; Land_area(sqkm), land area in square kilometres. b change in AIC scores as a function of the number of variables included in the model. c Variance inflation factor (VIF) test of multicollinearity among the 5 variables in the sub-model identified from all-subsets regression analysis. The x-axis refers to the VIF scores and the y-axis refers to the selected variables