The Vindication of Magnitude-Based Inference