Skip to main content

Table 2 Performance of the ensembled sex-aggregated XGBoost model on external datasets

From: Waist circumference prediction for epidemiological research using gradient boosted trees

Model: United Kingdom (National Diet and Nutrition Survey; 2000–2001) People’s Republic of China (China Health and Nutrition Survey; 2015)
Proportion of Dataset RMSE Bias Proportion of Dataset RMSE Bias
Overall 100 7.65 4.63 100 7.1 −1.7
Female 54.7 9.26 7.34 53.5 7.45 −2.02
 Asian 0.4 6.4 4.92 53.5 7.45 −2.02
 Black 1.2 4.81 2.85
 Hispanic 0.9 11.18 9.07
 Other/Mixed 1 11.39 8.18
 White 51.3 9.31 7.47
Male 45.3 5.41 1.73 46.5 6.67 −1.34
 Asian 0.5 4.13 0.34 46.5 6.67 −1.34
 Black 1 6.68 −3.08
 Hispanic 0.2 2.57 2.37
 Other/Mixed 0.8 4.21 1
 White 42.7 5.42 1.87
  1. Values represent sample-weighted mean bias or RMSE across the entire dataset
  2. RMSE root mean squared error