Figure 2
From: Pre-exposure cognitive performance variability is associated with severity of respiratory infection
Lumos session timing patterns corresponding to the 15 highest CPV vs shedding correlations. Left: Boxplots of association measures between CPV and shedding for which the cross-validated (leave-one-participant-out) CPV Pearson correlation coefficients are at least 0.69 (the lower endpoint of 95% CI of correlation in Fig. 1d). The measures of association are the Pearson correlation coefficient \(\rho\), the \(R^2\) of linear regression, and the AUC of logistic regression of titers onto CPV. The AUC measures the association of CPV with Low vs High shedding (0,1) labels, where Low and High denote shedding below or above the population median, respectively. The pattern heatmap at bottom indicates the corresponding timing patterns of Lumos sessions with Time 1 corresponding to the initial screening and Time 10 corresponding to the test right before exposure. The number of sessions in each pattern is denoted by T, fewer than \(T=7\) sessions significantly reduces the association. Right: relative frequency of inclusion of session times \(1, \ldots 10\) in (a) (summed over session patterns).
