Figure 5. Variable importance in random forest model.

Average importance of variables which predict participation/non-participation as calculated by the random forest classification model. The variables were classified into three categories which are (1) variables relating to child, delivery and pregnancy (Birth), (2) variables relating to neurodevelopment and behavior (Development), and (3) variables reflecting socioeconomic status (SES). Columns represent each age level of the longitudinal follow-up and are scaled to 100%.