In many cases yes. A simple example where this is not true is if your populations have different variances. T tests are no longer valid but heterogenous variances can be modeled with a Bayesian approach