Which metric measures, among instances flagged positive, how many are actually positive?

Get ready for the GARP Risk and AI Exam with flashcards and multiple choice questions. Each question comes with hints and explanations. Prepare for success!

Multiple Choice

Which metric measures, among instances flagged positive, how many are actually positive?

Precision is the metric that tells you, among all the instances your model labeled as positive, how many are actually positive. It answers the question: of the positives you predicted, what fraction are true positives? It’s computed as true positives divided by predicted positives (TP / (TP + FP)). This matters because a model can flag many positives, but if a lot of those are false alarms, precision is low. In risk contexts, a high precision means your positive alerts are reliable. Other metrics describe different ideas. Recall (sensitivity) asks what portion of all actual positives your model captured (TP / (TP + FN)); it’s about finding positives, not about the quality of the positive predictions. Accuracy looks at overall correctness, combining true positives and true negatives over all cases, which blends performance on positives and negatives. The ROC curve summarizes performance across different thresholds by plotting the trade-off between true positive rate and false positive rate, rather than giving a single measure of how reliable the positive predictions are.

Precision is the metric that tells you, among all the instances your model labeled as positive, how many are actually positive. It answers the question: of the positives you predicted, what fraction are true positives? It’s computed as true positives divided by predicted positives (TP / (TP + FP)). This matters because a model can flag many positives, but if a lot of those are false alarms, precision is low. In risk contexts, a high precision means your positive alerts are reliable.

Other metrics describe different ideas. Recall (sensitivity) asks what portion of all actual positives your model captured (TP / (TP + FN)); it’s about finding positives, not about the quality of the positive predictions. Accuracy looks at overall correctness, combining true positives and true negatives over all cases, which blends performance on positives and negatives. The ROC curve summarizes performance across different thresholds by plotting the trade-off between true positive rate and false positive rate, rather than giving a single measure of how reliable the positive predictions are.

More practice questions