Detection Positive Minutes, Hours or Days?
Large time unit – the drawback: if most units are positive the differences between sites are not seen. But : larger units also reduce the effect of variation in sensitivity between different loggers. (Dähne et al 2006).
Theory: If detections are random in time: DPM = fraction of 1minute units that are detection pos DPN = fraction of Nminute units that are detection pos n = minutes in Nminute units then DPN = 1 – (1 – DPM) n
Test on actual data:
This confirms the reduction in variation with larger time unit.
other statistics 'Time present' = the sum of train durations. This has the smallest units, and should correlate linearly with the density of echo-location activity. 'N of clicks' - this conflates behaviour with density as click rates in feeding buzzes are 10 or more times the mean rate. However buzzes may be rare … Some real data:
actual data: scale from 0.75
a note: Presence/Absence 'Presence/Absence' is often seen as qualitative or binary. In serial data there always is a time period within which presence/absence is determined. For small time periods, e.g. minutes or seconds, the % that are positive will scale nicely with density. So in time series data it’s effectively a quantitative measure …
Conclusion: To avoid a 'ceiling effect' the largest time unit that gives under 30% of units positive in all data sets to be analysed should be used. In ongoing studies a margin must be allowed for unknown future higher detection rates. 1min, 10min or 1hr are possible choices in most studies. Other considerations may apply in your project. For comparison with other studies DPM values should be given in reports or publications. For communication with wider audiences larger units such as Detection Positive Days per month or season will often work much better.
the end