Fig. 4.

Predicting conversion to psychosis based on semantic density in the original and shuffled samples. Individual points show (with a small amount of jitter) semantic densities of individual participants who either converted to schizophrenia (Probability = 1) or did not (Probability = 0). a Probability of conversion to psychosis given semantic density as estimated by binary logistic regression. b Probability of conversion to psychosis estimated by a model derived from the training data. c Probability of conversion to psychosis given semantic density estimated from randomly shuffling the language samples. After shuffling, conversion to schizophrenia was no longer predicted by semantic density