Biostatistics is more than t-tests and p-values
I have been doing this work for a long time. The Principles Of Biostatistics 2nd Edition by Manelau and Cornfield is one of those textbooks that actually stuck with me. I bought a used copy years ago because the library reference was checked out. It sat on my desk through multiple epidemiology courses and now it sits on my shelf and gets pulled out whenever I am reviewing study designs for grant proposals or teaching introductory methods. The book covers probability theory, distributions, estimation, hypothesis testing, regression, and experimental design. It is written for students who know calculus but have not necessarily taken a statistics course before. That matters because the pacing assumes mathematical maturity without assuming statistical literacy. You will see derivations, not just hand-waving. The examples are mostly from medicine and public health, which makes sense given the authors' backgrounds. I remember one specific problem I ran into while using it to prep for a board review. Chapter 8 on hypothesis testing has a section on type II error that uses a very dense derivation involving the normal distribution. I kept arriving at the wrong answer because I was treating the effect size as a proportion instead of a difference in means. The workaround was simple: I went back to the earlier chapter on sampling distributions and worked through the standard error formula from scratch. That gap in understanding was my fault, not the book's. The material is solid.
What the book does well
The treatment of probability is unusually thorough for a biostatistics text. Most books I have seen skim through Bayes' theorem and then move on. This one lingers. There are exercises that actually force you to think about conditional probability in a clinical context. The sections on experimental design, particularly randomized controlled trials, are clear and practical. You will find derivations of sample size formulas that make sense instead of formulas dropped from the sky. The regression chapter connects the dots between least squares and maximum likelihood. That is not trivial. Many programs teach these as separate topics. The book treats them as variations of the same idea, which is accurate and useful.
Where it falls short
The book is old. The second edition came out in the 1970s and even with reprints, the examples lag behind modern practice. Survival analysis is barely mentioned. Bootstrap methods do not exist in it. Multiple imputation is not covered. If you are reading this to prepare for work in clinical trials or epidemiology today, you will need supplemental material on these topics. The notation can be frustrating. The authors use Greek letters liberally and switch between capitalization conventions without always warning you. I lost time early on because I did not realize that when they wrote Y with a bar over it, they meant the sample mean of Y, not a different variable entirely. Once I got used to it, it was manageable.
Get the Full Details
How I use it
I keep it on hand for the probability and inference chapters. When I need a refresher on likelihood or a clean explanation of why we divide by n minus one for sample variance, this book works better than most alternatives. The exercises are challenging but fair. They are not computational drudgery. You actually have to think about what the answer means. For anyone self-studying biostatistics, I would pair this with a more modern text for the applied portions. Use this book for the foundation. Read something like Rosner or Kleinbaum for the current methods. The combination covers more ground than either book alone. You can find used copies on Amazon, AbeBooks, or eBay. New copies are sometimes available through academic resellers but the price is inflated. A well-worn paperback from a university bookstore sale is usually the best route. The content does not change significantly across editions for the core material. The first edition covers similar ground if you want to go cheaper.
If you are looking for a digital version, legitimate sources include university library portals and sometimes publisher sites. Avoid sketchy download links. The effort to locate a legal copy is worth it, and the quality of scanned PDFs from questionable sites varies a lot. The book will not teach you R or Python. That is not its purpose. It is a mathematically rigorous introduction to statistical reasoning for the health sciences. It does that job adequately, and for many readers, that is enough to build on.
