Associations of the probability of subclinical dengue infection with epidemic characteristics.*

*Epidemic characteristics at a given school, for a given epidemic year. Performed as individual logistic regression models with subclinical infection as the outcome variable and each epidemic characteristic as a single exposure variable, and incorporating random effects for the individual and the individual's school of attendance.**. Each exposure variable was aggregated across each school and for each year. The median represents the midpoint of these aggregated values.†A one-unit increase for proportions (incidence, proportion DENV-1 etc) was defined as a one-quartile increase in value. Quartiles were calculated based upon the range of values (e.g., if incidence had a range of 0–40%, the upper limit for calculating quartiles of incidence was 40%, not 100%). A one-unit increase in the number of serotypes in circulation compared 2 serotypes in circulation to 1 serotype in circulation, for example.