Reading the evidence
Observations are independent when knowing one does not change the relevant probability distribution of another under the model. Market events often share firms, dates or shocks and are not fully independent.
Make it click
A fresh coin toss does not use up outcomes, while drawing a card changes what remains in the deck. Both produce observations, but only one has the same unaffected next chance.
Now put that into markets
Companies reacting to one macro event can move together, and repeated observations from one company share structure. Treating them as independent overstates information.
Clustering, time-series models and block resampling address some dependence. Effective sample size is lower than raw count when correlation persists.
Map shared dates, entities, sources and overlapping return windows. Choose uncertainty methods that match the dependence structure.
No adjustment rescues a sample with too few independent clusters. Assuming independence because it simplifies software is not a method.
Takeaway
