[Dyac]

Fundamentally, you can't use the same data to both generate and validate/disprove a hypothesis.

Srgmenting and data dredging is fine provided you run a new test with fresh data to validate if there is a causal relationship in any correlations found.