post-hoc hypothesis testing

In standard hypothesis testing, the significance level $\alpha$ must be fixed prior to seeing any data. Choosing $\alpha$ in a data-dependent fashion can lead to p-hacking; see issues with p-values for more.

The goal of post-hoc hypothesis is to solve this problem and allow data to be data-dependent. This is an active area of research. Currently, the most promising solutions use e-values: see e-values enable post-hoc hypothesis testing. This approach involves reformulating error probabilities in terms of losses (post-hoc hypothesis testing with losses) (as in Neyman-Pearson paradigm with losses); how you feel about that will depend on you.