The SC2 prediction method explains that the sPTD vs Control, PPROM vs Control, should be independently predicted.
But we don't distinguish between sPTD, PPROM and Control, we try to predict the whole in a multinomial method.
Are you okay?
Created by DOHYANG KIM dohyang Both sPTD and PPROM groups have the gestational age at delivery (GADel) <37 weeks. The gestational age at blood draw (GA) are <36 weeks for all groups (Controls, sPTD and PPROM) .
When we assess prediction performance (AUC and AUPR) on the test set for the classification task Controls vs PPROM, only predicted probabilities of actual Controls and PPROM patients are used, so it does not matter what probabilities are predicted for patients in the sPTD group.
Similarly, when assessing the prediction performance for the classification task Controls vs sPTD, only predicted probabilities of actual Controls and sPTD patients are used, so it does not matter what probabilities are predicted for patients in the PPROM group.
I have a follow up question, since all PPROM in the training set have gestational age < 37, they are preterm. In the test set for predicting sPTD, if I classify a PPROM sample as sPTD, is the classification correct? for prediction, can I predict samples that are both sPTD and PPROM? I am not sure what the question is here. Indeed we do not ask to distinguish between sPTD and PPROM, bot only between each of these and controls.