Skip to content

“Physical Models of Living Systems”

Phil Nelson writes:

I’d like to alert you that my new textbook, “Physical Models of Living Systems,” has just been published. Among other things, this book is my attempt to bring Bayesian inference to undergraduates in any science or engineering major, and the course I teach from it has been enthusiastically received.

The book is intended for intermediate-level undergraduates. The only prerequisite for the course is first-year physics or something similar. Advanced appendices to each chapter make the book useful also for PhD students. There is almost no overlap with my prior book Biological Physics.

Rather than attempting an encyclopedic survey of biophysics, my aim has been to develop skills and frameworks that are essential to the practice of almost any science or engineering field, in the context of some life-science case studies.

I have quantitative and qualitative data on Penn students’ assessment of the usefulness of the class in their later work. This appears in the Instructor section of the above Web site.

Many of my students come to the course with no computer background, so I have also written the short booklet Student’s Guide to Physical Modeling with Matlab (with Tom Dodson), which is available free via the above web site. A parallel book, Student’s Guide to Physical Modeling with Python (with Jesse M. Kinder) will also be available soon. These resources are not specifically about life science applications.

This sounds great. And let me again recommend the classic How Animals Work, by Knut Schmidt-Nielsen.

P.S. We last encountered Nelson a couple years ago when answering his question, “What are some situations in which the classical approach (or a naive implementation of it, based on cookbook recipes) gives worse results than a Bayesian approach, results that actually impeded the science?” The question was surprisingly easy to answer. You might also want to check out the comment section there, because some of the commenters had some misconceptions that I tried to clarify.


  1. Martin says:

    It looks nice, but I will wait for the electronic format

  2. numeric says:

    I don’t really think much of your attractive birth selection example. One of the main criticisms of Bayesian methods is that you can get any result you want by a assuming the appropriate prior. You assume an appropriate (to you) prior and you negate the result in the article, but that is in essence proving the point of the criticism of Bayesian methods. I could simply assert that extraordinary claims require extraordinary evidence, which is the equivalent of setting the significance level of an unusual result to a higher level. In fact, I would think the two approaches (less prior belief in an effect, higher level of significance) would be almost isomorphic in most cases. Remember, .05 is a convention not a law handed down on tablets. The Higgs Boson experimenters demanded a level of .00001 (or .000001, I’m misremembering 10-5 versus 10-6).

    I think you are on much firmer ground when you defend Bayesian methods by arguing that they allow coherent ways of combining many sources of evidence and also that unlike classical corrections for multiple inferences, they shrink the effects of most differences rather than picking out a few wildly different estimators.

  3. DK says:

    Well, if the slides are anything to go by, then this book is not going to be useful to anyone.

Leave a Reply