Stata v17

Responses from the postal surveys were entered into Excel 2010 (Microsoft, Redmond, WA); online responses were downloaded into Excel 2010 format. Of the surveys completed on paper, responses were recorded as ‘missing’ for any unanswered questions., There were no missing responses on the online version. Data were transferred into Stata v17 [40 ] and merged into one dataset. To identify whether there was adherence to The Daily Mile taking place during curricular lessons at least 3 times a week, we grouped together the responses to question 12 and question 20 (see S2 Table). Descriptive data are reported as n and percentages. We used the Chi-squared test to assess differences in school characteristics between those responding to the survey vs those that did not, and between schools implementing The Daily Mile vs those that were not. All descriptive analyses were conducted using Stata v17 [40 ]. Statements provided in the optional comment box at the end of the survey were grouped by reference to principles and participation and non-participation in The Daily Mile.

Sensitivity analyses were conducted using two additional analytical samples: (1) those with a positive PCR test and (2) those with a positive PCR test and/or self-reported infection before entering the CIS. Furthermore, using the specifications of our main sample, a panel dataset was created where multiple participant visits and records were merged to create a dataset with one observation per month per person to allow for panel analysis that accounts for variation across time for time-varying variables, for example, ccupation, employment status (figure 1). We used multilevel mixed-methods generalised linear models, with a binomial link function for the outcome of having long-COVID symptoms; and a multilevel mixed-effects ordered logistic regression for the outcome of reduced function due to long-COVID and compared with the estimations of our analytical model.
All analyses were calculated by using STATA V.17.²²

Sensitivity analyses were conducted using two additional analytical samples: (1) those with a positive PCR test and (2) those with a positive PCR test and/or self-reported infection before entering the CIS. Furthermore, using the specifications of our main sample, a panel dataset was created where multiple participant visits and records were merged to create a dataset with one observation per month per person to allow for panel analysis that accounts for variation across time for time-varying variables, for example, ccupation, employment status (figure 1). We used multilevel mixed-methods generalised linear models, with a binomial link function for the outcome of having long-COVID symptoms; and a multilevel mixed-effects ordered logistic regression for the outcome of reduced function due to long-COVID and compared with the estimations of our analytical model.
All analyses were calculated by using STATA V.17.²²

Sensitivity analyses were conducted using two additional analytical samples: (1) those with a positive PCR test and (2) those with a positive PCR test and/or self-reported infection before entering the CIS. Furthermore, using the specifications of our main sample, a panel dataset was created where multiple participant visits and records were merged to create a dataset with one observation per month per person to allow for panel analysis that accounts for variation across time for time-varying variables, for example, ccupation, employment status (figure 1). We used multilevel mixed-methods generalised linear models, with a binomial link function for the outcome of having long-COVID symptoms; and a multilevel mixed-effects ordered logistic regression for the outcome of reduced function due to long-COVID and compared with the estimations of our analytical model.
All analyses were calculated by using STATA V.17.22

For statistical analysis, the predicted values of anti-RBD IgG and the percentage of inhibition measured by sVNT at the cut-off value for FRNT50 titers ≥20 and ≥40 were determined using non-linear regression analysis and performed on the log10 transformed data. The Spearman’s rank correlation between anti-RBD IgG, the percentage of inhibition measured by sVNT, and FRNT50 titers was determined using SPSS v23.0 (IBM Corp, Armonk, NY, USA). The r-square was calculated according to the non-linear equation using STATA v.17.0 software. A p-value < 0.05 was considered statistically significant.