Raising the contact threshold from 20 s to 5 min reduced raw recurrence and raised excess over a class and degree control in one 2013 school record

We tested how a daily contact duration threshold changes next-day recurrence. We used 188,508 sensor intervals from 327 students in nine classes at one French school during 2–6 December 2013. The control kept each next-day student's number of partners in every class. It changed partner identity by graph switches. With a 20 s threshold on both days, 3,825 of 9,138 source edge-days repeated (41.86%). With a 300 s threshold, 617 of 1,670 repeated (36.95%). Mean control recurrence fell from 32.59% to 10.60%. Excess above the control therefore rose from 9.26 to 26.35 percentage points. The increase occurred in all four transitions and remained after the shorter first day was removed. It was mainly within classes. In an exploratory check, any recorded partner return rose from 41.86% to 66.65% when the source threshold rose from 20 to 300 s and the target stayed at 20 s. Thus, less return above the same duration limit did not mean less partner return. The package contains source data, provenance, code, exact small-graph checks, results and figures. This is a threshold check and method replication in one record. Thresholds select different pair sets. The control does not separate friendship from seats, schedules or sensor absence. Its reference ranges are not confidence intervals across schools.

temporal networks · contact recurrence · duration thresholds · school contacts · degree-preserving randomization · SocioPatterns · reproducible reanalysis