Fig. 3: Earlier class start times were associated with shorter sleep opportunities.

a, Diurnal time courses of LMS logins are shown in students whose data were sorted by their first class of the day. LMS logins were compiled from 39,458 unique students. Coloured traces show data for school days and grey traces show data for non-school days in the same group of students. The vertical dotted line in each plot shows the start time of students’ first class. The grey boxes show LMS inactive periods on school nights when login activity fell below threshold (Methods). b, LMS login offset and onset values are shown for different class start times (coloured bars) and compared with data for non-school days (grey bars). LMS parameters were determined separately for five different semesters and the mean ± 95% CI is shown for each set of five values. The average duration of the LMS inactive period is indicated in each bar. LMS parameters (offset, onset, inactive period, midpoint of inactive period) for each semester and school start time are provided in Supplementary Table 3. c, Changes in LMS login behaviour are shown for different class start times relative to non-school days. The paired mean differences were calculated separately for each semester and the 95% CIs are shown for each set of five values. Grey traces show results for individual semesters. Linear mixed-effects models were used to test for differences in LMS login parameters for each class start time relative to non-school days (two-tailed t-tests; Supplementary Table 2). Multiple comparisons were performed using Tukey’s test and asterisks show pairwise comparisons that reached statistical significance (P < 0.05).