Title: A comparative study on the measures for the prediction accuracy of a survival model
Authors: Asanao Shimokawa - Tokyo University of Science (Japan) [presenting]
Etsuo Miyaoka - Tokyo University of Science (Japan)
Abstract: Evaluating the prediction accuracy of a survival analysis model is one of the important problems in practical applications. Also, these evaluations are used for assessment purpose of an interesting risk score ability in the model. Several different measures have been proposed for this purpose: discrimination based approaches like Harrell's C-index and modification versions of it, sensitivity and specificity based approaches or ROC and AUC based approaches in a similar sense, and explained variation or loss function based approaches. We compare the performances of these evaluation measures of the Cox proportional hazard model through simulation studies under several situations such as incorrect the covariates or interactions that should be included in the model. In addition to this, we show the results of applying these measures to the model obtained from an actual data.