My name is Russell J. Bowater (aka The naked statistician) and welcome to my How Should We Do Statistics website.
I am an independent scientist and statistical consultant/adviser based in the city of Oaxaca in the south of Mexico, and honorary researcher / research fellow in various institutions. I have lived for extended periods of time in England (where I was born and brought up), Italy (for 4 years) and Mexico (for more than a decade). If you are interested, my academic and work history can be found here and a list of (most!) of my publications can be found here, while talks that I have recently given can be found here. Please feel free to contact me via email or my Twitter/X account.
This crisis has been caused by a perceived failure of conventional statistical methods to serve the needs of science, e.g. the fallout from the replication crisis in psychological research, and the incapacity of leading statisticians to reach a consensus on what are the most appropriate statistical methods, theories or schools of thought to apply even to the most basic problems in statistical inference. Perhaps the clearest symptom of the continued impact of this crisis are the hostile debates between supporters of the frequentist school of inference (frequentists) and supporters of the Bayesian school of inference (Bayesians) about how, in general, statistical inference should be carried out.
The foundations of statistics crisis will not be resolved until we have a widely accepted solution to the fundamental problem of statistical inference (FPSI), which may be described as the problem of finding a satisfactory completely general framework for performing statistical inference.
Even though this problem has been heavily overlooked, many argue that we should overlook it some more! Their reasoning is as follows:
1) It was already been solved. The solution is Bayesian inference.
To maintain this position, you would need to be incredibly thick-skinned to the obvious criticisms that can be made of Bayesian inference.
2) The framework that would represent a solution to this problem would be cumbersome to use and therefore would be remote from the needs of practical statistics.
It is difficult to evaluate the adequacy of practical statistical methods without a reference point for comparison. A solution to the FPSI would provide the ideal reference point.
3) The problem has not been solved up to now and therefore why waste time in trying to solve a problem that is probably unsolvable.
This, of course, has been said about many problems in mathematics and science which were then later solved, e.g. Fermat's last theorem (which was of course solved by Andrew Wiles in 1995). Also, I have made sufficient progress in addressing the FPSI that I can now see what a solution to this problem would look like. I would like to collaborate with other statisticians who are interested in solving the FPSI but at the moment, I do not see anyone else who is moving towards a solution and so I will continue to work alone on this problem.
This website aims to publicise my research work and my progress, as well as the progress of others, in solving the fundamental problem of statistical inference.