According to Google Analytics, my top content of 2010 was
O'Brien Fleming Designs In Practice
My O'Brien Fleming Design Is Not Same As Your O'Brien-Fleming Design
Matrix Square Roots In R
Barnard's Exact Test: Test That Ought To Be Used More
Perverse Incentives In Clinical Trials
Bayesian Information Criterion
Review Of Jim Albert's Bayesian Computing in R
Note that views of the front page are excluded, but as that content rotates it is hard to interpret that. I have put some easy feedback mechanisms for each post (including the rating scale), but these have not returned very much information.
None of this surprises me. It seems like O'Brien Fleming designs are the top content just about any time I check, and I think this is because adaptive trials are becoming more popular. The FDA's draft guidance on the subject was released, and I guess O'Brien-Fleming designs are the most popular type of adaptive trial at least in the confirmatory stages of drug development.
I'm still surprised that matrix square roots in R is next, because the Denman-Beavers algorithm is fairly inefficient unless you need the square root and its inverse. Some method based on eigenvalues or singular value decomposition is probably better, but I guess occasionally you run across the odd matrix that has a rank 2 or higher subspace in its eigendecomposition. I'm glad to see Barnard's test getting a little exposure if at least to challenged the flawed overapplication of Fisher's exact test. I'm surprised Perverse Incentives and NNT did not rank higher, as those topics I think are of wider interest, but there you go. The BIC is a fairly specialized topic, and while model selection is of wide interest the methods I used that involved the BIC are rather specialized.
As a note, I write the review of Jim Albert's book before the second edition came out. I have the second edition now, and I should probably get around to updating the review.
As any reader of this blog might be able to tell, my focus is probably going to change a bit as I get more comfortable in the post-marketing world of drug development. The issues change, the quality and kinds of data change, and so my day-to-day issues will change. We'll see how it unfolds.
Of course, I'm open to suggestions as well.