In one 2013 school week, wider contact spans had a 6.3-point higher next-day rate after exact student and count matches

We tested whether the time span of daily contacts gives information about contact on the next day. We used 188,508 sensor records from 327 students at one French school on 2 to 6 December 2013. A wide span meant at least one hour between the first and last contact. Exact matches held the focal student, partner class, day, and daily contact count fixed. The matched next-day rates were 49.4% for wide spans and 43.1% for shorter spans, a difference of 6.3 percentage points. All four daily differences were positive. Matching kept 12.6% of endpoint records and 2.9% of contact time. The result mainly describes brief contacts. In a fixed forward test on two later transitions, adding the span indicator reduced pooled log loss from 0.609599 to 0.607463. This was a small gain. Shared schedules, observation time, and pair preferences can contribute to the relation. The study does not show a causal effect or transfer to other schools. The package preserves the PLOS input, source provenance, code, figures, exact matches, fitted models, predictions, and reproduction instructions.

temporal networks · contact span · contact recurrence · school contacts · exact matching · forward prediction