Glossary terms for 'K'
| K-fold cross validation | A method of validating an estimate of a model's performance that may have been inflated by overfitting. For example, in a 10-fold cross validation, the sample would be divided into ten groups. Ten separate models would be derived, each time leaving out one of the tenths. Estimates of model performance are based on the predicted and observed values in the 10 different left-out groups. |
| Kappa | A statistical term that measures the degree to which two (or more) observers agree whether a phenomenon occurred, beyond that expected from the marginals. Varies from –1 (perfect disagreement) to 1 (perfect agreement). For example, the kappa comparing how well two pathologists agreed about the presence of cirrhosis in a sample of liver biopsy specimens was 0.55. |
Glossary material from Hulley SB et al. Designing Clinical Research, 5th ed. Philadelphia, Lippincott Williams & Wilkins, 2022.