Bayesian Statistics

Classical Statistics, the stats we all learned in Stats 101 in university, was developed to support science.  Like science, classical statistics is all about rejecting (never accepting, even though it is often taught in stats classes and described in many a stats textbook erroneously as " accepting") a hypothesis.  With classical statistics, we only continue to consider a hypothesis as possibly true because we have not found enough evidence to reject it as false.  Making a decision in engineering to use a particular technology in a design only because we have not found enough evidence to reject it is pretty weak.  In systems engineering, deciding to accept a subsystem to use in a manned space mission only because we have not found enough evidence that it might fail can be disastrous.  In project management, making a decision to take a certain path with a subcontractor only because we have not found enough evidence to reject them can lead to project failure.  Classical statistics was not really intended for engineering, systems engineering, or project management.  Plus, there are a lot of assumptions necessary for any results produced by classical statistics that are very rarely discussed, and are very rarely valid for real world problems.

Bayesian statistics, really just a good and thorough application of probability, was however intended to assist with decision making, and that includes decision making in engineering, systems engineering, and project management.  Bayesian statistics is all about determining how sure we can be that a hypothesis is actually true based on the data and information we do actually have.  This is really what we need in engineering, systems engineering, and project management, not making decisions to use an alternative only because we could not reject it.  In fact, all of the statistics in all of the decision theory and risk foundational textbooks has been Bayesian statistics.  

The question then arises as to if this is the case, why have we not been using Bayesian statistics in engineering, systems engineering, and project management for the past 70 or so years?  The answer is quite simple:  solutions derived using Bayesian statistics for real world problems are almost never analytically tractable nor even solvable with ordinary numerical methods.  Around the mid-1990's however, new numerical methods were developed called Markov Chain Monte Carlo methods that enabled quantitative solutions developed using Bayesian statistics for real world problems.

Attwater consultants are recognized internationally as experts in solving problems and decision making using Bayesian statistics via Markov Chain Monte Carlo methods.  Numerous of our published papers used Bayesian statistics and Markov Chain Monte Carlo to solve problems that were heretofore not satisfactorily solved.  We have developed methods for using Bayesian statistics to avoid any questionable assumptions, which leads to the ultimate in unbiased results.  We use Bayesian statistics exclusively, and have achieved quick, easy, and comfortable decision making for our clients 100% of the time.  We teach Bayesian statistics in our education and training offerings as well, and can teach your staff how to make much better engineering, systems engineering, and project management decisions.


Attwater Consultants provide unique and powerful skills for good decision making in engineering, systems engineering, and project management through use of Bayesian statistics and Markov Chain Monte Carlo methods.   Contact Us for more information on how an Attwater Consultant can help fine tune your Projects into assured successes.  We selectively accept short term and long term assignments all over the world; language used is English.