One way of generalising thresholds for large datasets is to use a
percentile, for example the 90th percentile (DuMouchel, 1983), or the
square root of n, where n
is the number of data (Ferreira et al., 2003). Threshold selection has been the subject of wide
research, and an overview of the different options is given by Scarrott and MacDonald (2012).
3.3 Estimation of the distribution
/media/vedurstofan-utgafa-2020/VI_2020_008.pdf