I am trying to determine the MCID of a PROM using a anchor based method (ROC). I have the following dataset.
The dataset has 800 rows. I'm using python, and I did the following:
- Train/test split the data
- Fit a Logistic model
- Predicted probability of 1
- Plotted ROC using y_test and predictions
My AUC is 0.8, which is a good value for this type of methodology and context. I calculated the Youden Index which is 0.87. Now I'm stuck: How do I know the Delta_PROM threshold using this value?