Is there any university program that offers an introductory statistical methods class that is bayesian -instead- of a frequentist one?

Surely every intro to stats class in every program in the world, whether agronomy or economics or psychology or a service class by the stats department, is primarily frequentist based.

And just as surely there are many (but not all) programs where later in the curriculum there are some specialty courses (requiring a few stats pre-reqs) that are titled 'Intro to Bayesian Statistics'.

But are there at least some places somewhere that offer a bayesian class as the intro statistical methods class? (either as the only intro to stats -or- simply as an alternative first class to a frequentist intro)

I would count an intro stats class that teaches bayesian methods in addition to frequentist ones.

If such programs are not as rare as I imagine, then is there any indication of percentage?

I am wondering how easy it is to have a bayesian methods course without first having a frequentist methods course.

  • $\begingroup$ As suggested at the original question over at stats.SE, McElreath’s textbook "Statistical Rethinking: A Bayesian Course with Examples in R and Stan' would be the start-from-scratch Bayesian methods text to use. $\endgroup$
    – Mitch
    Commented Oct 7, 2020 at 15:33
  • $\begingroup$ I'm finding (and this was the motivation for the original question) that the multitudinous post-grad/practical Masters data science programs are descriptive stats/probability, then NHST/p-values, then OLS and done. Bayesian methods don't even exist. Of course there's an ML component but that's not Bayesian, just random assumption-less regression methods. $\endgroup$
    – Mitch
    Commented Oct 11, 2020 at 17:30
  • $\begingroup$ OLS (and least-squares more generally) can be used in a Bayesian way: in that class I taught at Plymouth University, I used to discuss algebraic formulae that could be applied to the typical output values provided by least-squares fitting software, to obtain the leading-order Laplace's method approximations to the expectation and the marginal standard deviations of the posterior distribution over parameters, and to the marginal likelihood for the model; I also used to present some Gnuplot scripts that implemented those formulae. $\endgroup$ Commented Oct 20, 2020 at 11:00

1 Answer 1


I taught an introductory statistical methods class that was wholly Bayesian a few times, on the Mechanical, Marine, and Materials Engineering programmes at the University of Plymouth, but I'm not sure I'd consider it a great success (the class had no summative assessment of its own, being assessed through the "analysis of evidence" element of the undergraduate dissertation, and the vast majority of students found non-statistical ways to demonstrate their "analysis of evidence" skills and didn't engage with the class at all).

In my own student life (Natural Sciences programme at the University of Cambridge, with main specialism Experimental and Theoretical Physics), the introduction to statistical methods was through a second-year undergraduate course called "Experimental methods", which was neutral between Bayesian and frequentist (the statistical elements of the class were based on a parameter estimation method which can be justified either from a Bayesian or a frequentist perspective, and a model assessment method which can be viewed either as a very rough approximation to a Bayesian model comparison or as a very rough approximation to a frequentist significance test - the class never referred to either the Bayesian or the frequentist foundational aspects), followed up by a class in the master's year called "Information theory, pattern recognition and neural networks", which was the course from which David MacKay's book Information theory, inference and learning algorithms grew, and which was thoroughly Bayesian.

  • $\begingroup$ Nice. So this was for engineering undergraduate degrees then? (sorry, I only know the American degree system labels). $\endgroup$
    – Mitch
    Commented Oct 11, 2020 at 17:24
  • $\begingroup$ @Mitch Yes, undergraduate engineering degrees. $\endgroup$ Commented Oct 11, 2020 at 20:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.