Contacts between classes repeated more often than a model with fixed daily partner counts expected in three campaigns at one school

We tested whether contacts between classes repeat more often than a model that keeps each student's daily number of partners in each other class. We used three public campaigns at one school in Marseilles. The main test used 2013. It gave 266 repeat counts across four successive day pairs. The model mean was 153.21, for an observed-to-model ratio of 1.74. A simpler model that kept class totals alone gave a mean of 37.13 and a ratio of 7.16. At a threshold of one minute on each day, the fixed-partner ratio was 1.96. The main-threshold ratios were 1.66 in 2011 and 2.11 in 2012. All 13 day-pair counts exceeded their model means. We checked the sampler by full enumeration on small examples, longer runs, and a separate two-edge switch implementation. Source checksums, direct overlap counts, and repeat-run arrays passed. Prior work already showed contact stability with a related degree control. This contribution gives a binary count replication and a 2013 extension with code, source provenance, aggregate results, model samples, and figures. The data cover one school. The model does not keep room use or schedules within a day. The result does not establish friendship, cause, or infection risk.

school contacts · temporal networks · contact repetition · conditional null model · Curveball · reproducible research