This paper introduces the nonparametric predictive inference approach for reproducibility of likelihood ratio tests. The general idea of this approach is outlined for tests between two simple hypotheses, followed by an investigation of reproducibility for tests between two beta distributions. The paper reports on the first steps of a wider research programme towards tests involving composite hypotheses and substantial computational challenges.
- Beta distribution
- Likelihood ratio test
- Lower and upper probabilities
- Nonparametric predictive inference
- Reproducibility probability