A member of our local IRB forwarded a press release announcing new software that would revolutionize the conduct of clinical trials. The company is called Analytical Edge and they offer a software program called Pure Likelihood. There is only a limited amount of information at the website:
But the web page of one of the founders of the company, Jeffrey Blume, includes some talks and papers about the topic. The basic thrust is that the p-value is not a good tool for making decisions about whether a drug is safe and/or efficacious, and it should be replaced with the likelihood ratio. The best paper
- Likelihood methods for measuring statistical evidence. J. D. Blume. Stat Med 2002: 21(17); 2563-99. [Medline] [Abstract] [PDF] (Model, Likelihood)
is well worth reading. The likelihood ratio is much more nimble at capturing information appropriate for interim analyses, for example.
Dr. Blume is not the first to criticize the p-value, and some interesting links if you want to explore this further are listed below.
- Special Issue: Statistical Significance Testing. Dennis Roberts, Penn State University. Accessed on 2003-03-20. “The Fall 1998 Issue of Research in the Schools, was a special full issue on Statistical Significance Testing. This issue contained 6 primary papers and 3 follow up comments. The Editors and Publishers of Research in the Schools agreed to have this issue put in a web format." roberts.ed.psu.edu/users/droberts/sigtest.htm
- 326 Articles/Books Questioning the Indiscriminate Use of Statistical Hypothesis Tests in Observational Studies. William L. Thompson. Accessed on 2003-03-19. “The following list of citations is the result of an extensive, but by no means complete, search for articles/book chapters that address the problems associated with applying statistical hypothesis tests (a.k.a. null hypothesis tests, null hypothesis statistical tests, hypothesis tests, statistical significance tests, significance tests, etc.) to observational studies." www.cnr.colostate.edu/~anderson/thompson1.html