论文部分内容阅读
Recent advancement in technology has lead to much progress in disease prognosis.With dramatically increased availability of ncw predictive markers, it is now possible to make a more accurate prognosis by combining information from multiple markers.Although combining information from several markers may lead to improved prog nostic accnracy, it may not be cost effective to obtain all marker measurements for all patients.In many clinical applications, prognostic modalities with higher accuracy are often expensive and/or invasive.It is thus crucial to assess the incremental value of ncw markcrs.Such assessment is typically made by averaging across patients in the entire study population.However, an overall improvement does not justify mea suring the new markcrs in all patients when there is cost associated with measuring the markers.A more practical strategy is to utilize conventional markers to decide whether the new markers are needed for improving prediction of outcomes.In this research, we propose robust non-parametric inference procedures for the incrcmental values of new markers across various subgroups of patients indexed by the conven tional markers.The proposed methods employ working models to approximate the associations between the predictors and the outcome but derive non-parametric es timates of the incremental values across various sub-populations without requiring the working models to hold.The resulting point and interval estimates can be quite useful for medical decision makers seeking to balance the predictive or diagnostic value of new markers against their associated cost and risk.