Need to determine lead or lag and magnitude of relationship between behavioural indicators and (changes in) transmission.
Consider the correlation metric used. Pearson's rho considers linear correlation, whereas others (such as distance correlation) captures non-linear relationships and general monotonicity.
Explore existing methods for calculating this, such as whether to correlate entire time series or bootstrap contiguous time windows.
Remove underlying AR(2) terms? Correlate raw or transformed data streams? How to inform lag structures e.g. how consistent does the lag need to be and what if there is no consistent lag? How does this relate to the model structure used?
Need to determine lead or lag and magnitude of relationship between behavioural indicators and (changes in) transmission.
Consider the correlation metric used. Pearson's rho considers linear correlation, whereas others (such as distance correlation) captures non-linear relationships and general monotonicity.
Explore existing methods for calculating this, such as whether to correlate entire time series or bootstrap contiguous time windows.
Remove underlying AR(2) terms? Correlate raw or transformed data streams? How to inform lag structures e.g. how consistent does the lag need to be and what if there is no consistent lag? How does this relate to the model structure used?