New statistical ways for analyzing sleep data in clinical and preclinical studies

Sleep-EEG data in human and animal studies are generally characterized by a multitude of parameters gained directly or derived indirectly from aggregations and transformations. Plenty variables mean the availability of more information leading to enhanced quality in description and understanding. However, the understanding of mechanisms and causalities of the observed phenomena is a difficult task. Especially in empirical sciences, where the observed phenomena are characterized by a great variability both in time and space dimension it is a challenge to establish omnipotent laws. There is a need for statistical models approaching better the reality.We searched new statistical approaches for a deeper insight into the structure and dynamic behavior of the sleep processes: time shifted association analyses, semiparametric multinomial and time-dependent regression, intraclass-correlations and genetic variances of time-varying sleep variables, stability tests, multiple graphs. Almost all these methods are meanwhile implemented as separate modules into the Event History Analysis program, which is applied in sleep research. It helps to perform the aforementioned analyses and many other data evaluations and presentations in a very comfortable and easy manner. Supported by the DFG (YA 12/1–2) Kalus et al, Am J Physiology 2009,296,R 1216–27