The authors wish to note the following: “It has been brought to our attention that our computer code for calculating cosinor statistics inadvertently used the frequency in degrees and not in radians, leading to an underestimate of the frequency of true positives. We have corrected the code available on github (https://github.com/murthylab/noIPIcycles) and recalculated the statistics that appeared in Fig. S1. The new Fig. S1 illustrates that true positives occur more frequently than positives in the shuffled data. However, cosinor still demonstrates a substantial false positive rate (the positives present in shuffled data). In addition, the positives for natural song are dispersed across a wide range of period (20-150 s), as we illustrated previously using other methods to estimate periodicity, and do not tend to cluster near 60 s, which is the approximate periodicity that has been implicated previously for interpulse interval cycles (i.e., K-H cycles). In fact, half of the periodicities detected are faster than 40 s. As we discussed in the original paper, rare periodicity scattered across a wide frequency range is more consistent with a model in which male flies modulate their interpulse interval in response to social cues than it is with a model in which an internal clock drives regular periodicity. This corrected analysis therefore still supports our earlier conclusion that there is no statistical evidence that flies sing pulse song with interpulse intervals that cycle with a regular periodicity.” As a result of this error, Fig. S1 in the SI Appendix and its legend appeared incorrectly. The SI Appendix has been corrected online.
All Science Journal Classification (ASJC) codes