All evidence about $\theta$ is contained in the likelihood function given the observed data.

The sampling distribution of the data is more important than the observed likelihood.

Only frequentist methods are valid.

Bayesian priors should always be non-informative.