出版社：SISSA, Scuola Internazionale Superiore di Studi Avanzati
摘要：The multiplicative model of Uttley et al. (2005) for the X-ray emission from X-ray binaries accounts well for the observed global characteristics of the time series such as the linear flux-rms relationship and the lognormal distribution of the fluxes. However, models that are fitted to the bulk of the data may not provide satisfactory models for the extreme events in the series: these might be produced by other unrecognized processes. As the model of Uttley et al. (2005) implies a well-determined structure of events above high thresholds, we applied methods from extremevalue statistics to check whether the same process can be responsible for the extremes in the data. We first considered the limiting distribution of large events as expected from the model, and compared to that of a model fitted to data from Cygnus X-1. Next, the clustering characteristics of the observed large events was compared to what is expected. Both comparisons show that the multiplicative model can indeed generate the observed extreme events, though there are indications of nonstationarity in the baseline level of the process and also of short periods where the observed extreme-value characteristics could be incompatible with the model implications.
