Fig. 5

Partial dependance plots of a selection of top ranked categorical variables within a random forest model to predict self-perceived general health status. Legend: The following variables from the complete 2016 random forest model (n = 244,557, variables n = 91) are presented: a “Control of own life” (VI rank: 1), b “Guideline physical activity (VI rank: 3), c “General loneliness” (VI rank: 4), d “Making ends meet” (VI rank: 5), e “Guideline muscle- and bone-strengthening exercise” (VI rank: 6), f “Gainful Employment” (VI rank: 7), g “Incapacitated” (VI rank: 8), and h “Emotional loneliness” (VI rank: 11). Dashed line represents the probability of poor SPGH status in the dataset without taking any exposures into account (0. 041)