Project Details
Projekt
Regression and multiple time scales in multi-state models (M: Multi-state)
Applicant
Professor Dr. Martin Schumacher
Subject Area
Epidemiology and Medical Biometry/Statistics
Term
from 2004 to 2011
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 5470786
To comprehensively judge the course of a patient's disease, different events and their timing need to be taken into account. Nonparametric event history analysis allows the statistical treatment of such data in many clinical situations. Applications, however, are comparatively scarce and mostly concentrate on transition intensities. Frequently, clinical questions are too complex to be answered by looking at transition intensities alone; rather, so-called summary measures are needed. In addition, they have the benefit of being more ready to interpret. In this project, we will focus on analyzing the effect of intermediate events in terms of summary measures. For some of them, like expected change in length of hospital stay due to an intermediate event (some complication, say), admissible estimators, their asymptotic expansion and testing procedures are virtually unknown. Modeling the influence of covariates on such summary measures is an active field of research. Summary measures themselves - 'Which one should we use?' - are the subject of vital discussions among both statisticians and clinicians. Our mission is to bridge the gap between the statistical theory and clinical epidemiology, supplying better answers to clinical questions that arose in preceding cooperations and to specifically address methodological problems characteristic to an advanced analysis of intermediate events. We will do so in tight cooperation with other projects of the research unit, putting special emphasis on prognostic factor modeling in time, causal inference and performing and assessing dynamic predictions. Data sets will be shared and additional methodological support will be offered by our project.
DFG Programme
Research Units
