2021/08/08 by Nicholas Adams, Adams, Nicholas
Computer Science · Mathematics · Physics and Astronomy · #Bayesian Modeling and Causal Inference #FOS: Computer and information sciences #Methodology (stat.ME) #Probability and Statistical Research #Statistical Mechanics and Entropy
paper · pdf · doi:10.48550/arxiv.2108.03544
openalex publication_date 2021/08/08 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
The one-sided P-value has a long history stretching at least as far back as Laplace (1812) but has in recent times been mostly supplanted by the two-sided P-value. We present justification for a bijective relationship between the one-sided P-value and a likelihood ratio based on maximum likelihood, a relationship that cannot be demonstrated for the two-sided P-value. A number of criticisms of P-values are discussed and it is shown that many of these criticisms are not justified when a likelihood ratio interpretation of a one-sided P-value is employed. Converting a one-sided P-value to a likelihood ratio provides the advantages of the likelihood evidential paradigm.