Likelihood

Rating:
Summary: An interesting mix
Review: Reasoning in terms of likelihoods is completely routine these days. In many areas - string matching, signal decoding, model-building - it is used without any special mention, like any other fundamental. It seems hard to remember that, not too long ago, likelihood and Bayesian techniques were the topics of impassioned debate - somtimes on different sides of the debate. This little book dates back to that era.

The book addresses the use of likelihood in a number of familiar applications (parameter estimation, etc). The examples are numerous and clear. I find more recent writings to be more directly applicable, though.

The real value of this book, for me, is the historical perspective that the author brings to the discussion. Early in his career, he worked with Fisher - one of the founders of modern statistics. Edwards describes how likelihood-based analysis grew out of that tradition. He also restates some of the philosophical objections to Bayesian analysis, especially questioning the use of priors.

I didn't come away from this book with any new analytic skills, though I suppose I could have mined the text more deeply. I did come away impressed by how much applied math and probability modeling have changed just since the 70s.