Fig. 3From: Trends in the ability of socioeconomic position to predict individual body mass index: an analysis of repeated cross-sectional data, 1991–2019Predictive accuracy of random forest models predicting individuals’ BMI by survey year. a Mean absolute error of model predictions by model (i.e. average difference between predicted and observed BMI; baseline prediction uses sample mean, and other estimates are random forest models including stated covariates). Higher values are indicative of less accurate prediction. b Percentage reduction in prediction error when further including educational attainment, social class, or IMD in the random forest model (compared to the model including age and sex). c Incremental R2 when further including educational attainment, social class, or IMD in the random forest model (compared to the model including age and sex). d Probability of superiority by modelBack to article page