For which type of outcome cases should a high predicted probability be observed in binary predictions?

In binary predictions, a high predicted probability should be observed for primary outcome cases. This is because primary outcomes represent the main result of interest in a study or model. When performing binary classification, the goal is to accurately identify instances that reflect the primary outcome, typically denoted as the event of interest, such as success or failure.

A high predicted probability indicates a strong model confidence that a case belongs to the primary outcome category. This is particularly important in contexts like medical diagnoses, fraud detection, or any other area where distinguishing between the occurrence or non-occurrence of the primary outcome is crucial for effective decision-making and actions.

In contrast, secondary or minor outcome cases do not carry the same level of importance in the context of the analysis, and thus would not typically show high predicted probabilities relative to the primary outcomes. Miscalculated cases refer to instances where the prediction is inaccurate, which would also not lead to high probabilities for the actual outcome. Therefore, the focus on high predicted probabilities should always be on the primary outcome cases, confirming their significance in the model's performance.