Which statement correctly defines Specificity?

Which statement correctly defines Specificity?

• A. Specificity is the probability the model predicts a negative outcome for an observation when in fact the observation is negative.
• B. Specificity is the probability the model predicts a positive outcome for an observation when in fact the observation is positive.
• C. The ROC Curve plots the false positive rate against the true positive rate.
• D. AICc should be used when the sample size is under 200.

Answer: (A)

Specificity is the probability that the model assigns a negative prediction when the actual observation is negative. In other words, among all actual negatives, what fraction does the model correctly identify as negative? This is the true negative rate, calculated as true negatives divided by the total number of actual negatives (true negatives plus false positives). The statement describes exactly this conditional probability, so it correctly defines specificity. For context, the complementary concept is sensitivity, which looks at actual positives and the chance the model predicts positive. The ROC curve relates to trade-offs between false positive rate and true positive rate, not the definition of specificity itself. AICc is a model-selection criterion for small samples, unrelated to specificity.

Specificity is the probability that the model assigns a negative prediction when the actual observation is negative. In other words, among all actual negatives, what fraction does the model correctly identify as negative? This is the true negative rate, calculated as true negatives divided by the total number of actual negatives (true negatives plus false positives). The statement describes exactly this conditional probability, so it correctly defines specificity.

For context, the complementary concept is sensitivity, which looks at actual positives and the chance the model predicts positive. The ROC curve relates to trade-offs between false positive rate and true positive rate, not the definition of specificity itself. AICc is a model-selection criterion for small samples, unrelated to specificity.