Fig. 5From: Mixture density networks for the indirect estimation of reference intervalsReference intervals for hemoglobin concentration. Top row: Hemoglobin concentration of girls with estimated quantiles \(Q_p\) for the healthy component where \(p \in (0.025,0.1,0.25,0.75,0.9,0.975)\). Displayed results show the solutions with minimum loss after 100 repeated initializations. Dashed black lines represent the EM-algorithm, the slightly finer solid lines discrete results from an alternative approach not based on latent class regression. Bottom row: Estimated mixture component weight of the healthy componentBack to article page