Sequential Test for Practical Significance: Truncated Mixture Sequential Probability Ratio Test

Abstract

We present a sequential testing method to identify a practically significant effect. We build on the existing mixture sequential probability ratio test (mSPRT) that can sequentially test for a non-zero treatment effect by using a truncated mixing distribution to differentiate between effects that are large enough to merit a real world action versus that are non-zero but too small to merit a real world action. We verify the Type-I error control of our method theoretically and empirically. We also extend this idea to sequentially test for one-sided practical significance such as non-inferiority testing, and show that we may still effectively control the Type-I error rate.