Teil C - Annex 
69 
al. 2003, Tougaard et al. 2009), the first waiting time after end of pile driving activities must be 
compared to a random sample of the uninfluenced waiting times; the randomized sample size 
(n) must be identical. This has to be effected by randomised selection of dates/times and re 
lated waiting times from time intervals that are uninfluenced by pile driving and feature similar 
seasonal patterns or from longer intermissions in the pile driving activity, during which a natu 
ral distribution of harbour porpoises that is uninfluenced by pile driving may be assumed (Tou 
gaard et al. 2009). The Bus Paradox is not eliminated by randomised selection of numbered 
waiting times in the randomised sample, but exclusively by randomised selection of a point in 
time and selection of the waiting time associated with that point in time. In this case, waiting 
times should be used only once (selection without putting back/”Jackknife”). All other re 
viewed waiting times after pile driving are independent of the “Bus Paradox” and are com 
pared with the total number of uninfluenced waiting times. 
Non-parametric standard test procedures (e. g. Mann-Whitney U-Test) lend themselves as 
mean value comparisons between individual groups (e. g. first “waiting time” after pile driving 
vs uninfluenced “waiting time”, separated into distance classes). Further analyses allowing for 
the investigation of influence of other parameters may be carried out by using a generalised 
linear model (GLM), which can be extended to a mixed model (GLMM) by addition of random 
factors (e. g. POD station).