Fig. 5From: A machine learning approach to small area estimation: predicting the health, housing and well-being of the population of NetherlandsXGBoost SHAP values for “drinker” indicator.. Because SHAP values explain each individual’s prediction as a sum of the contribution of their features, we calculate the average SHAP value of these individuals for a given feature value. These have intuitive interpretations. Positive contributions to drinking are: age in early 20s, sex is man, being divorced, higher socioeconomic status. Negative contributions: being retired, sex is woman, ethnic backgrounds where Islam is the main religion, larger household size, income and assets around the lowest 25%Back to article page