Association persistence (stable / variable)
A label saying whether the observed association held across shorter sub-periods or appeared only once.
A single whole-window result is easy to mistake for a lasting trait even when it is the effect of one turbulent week. So we split the pair's shared days into three consecutive sub-periods and compute the association separately in each. If every computable sub-period lands in the same descriptive band as the full-window result, we mark the observation "held across sub-periods" . If even one falls into a different band, it is "appeared only in part of the window" . When the sub-periods are too short to compute anything we do not guess: the label stays "not enough data to assess persistence" . The label says nothing about the future, it describes only how repeatable the observation was inside the analysed window.
In plain words
We check whether what shows up across the whole window also shows up in its three chunks. If it does, the observation was repeatable. If not, it is more likely a one-off coincidence and should not be treated as a fixed trait.
Example
A pair at r = +0.04 over 90 days gets the label "appeared only in part of the window", because the first sub-period came out at +0.05, the second at +0.11, but the third at +0.42, a different band already. Another pair at r = +0.88 stays above 0.85 in all three sub-periods, so it gets "held across sub-periods" .
Also known as: stabilność korelacji, stability label