All evidence about $\theta$ is contained in the likelihood function $L(\theta; x)$.

Frequentist $p$-values are the most robust measure of evidence.

The prior distribution is essential for any valid statistical test.

Large sample properties are irrelevant for small datasets.