Friday

Reading

  • Arxiv has several articles on the EM-algorithm from J. Statistical Science unknown, unknown.  What makes this distinct from applying any traditional optimization routine (Nelder-Mead, simulated annealing, etc) to the likelihood function?  Just in that it determines the probability distribution as well as maximum?

  • UPDATE ON EM: brief discussion with Graham, Peter, Yaniv on this, now my impression is: Not necessarily get the distribution, but doesn’t just generically take a function and minimize it, but may not calculate the likelihood at each step.  Handles missing values, simulate the data.  We’ll take a closer look at this in our Algorithms discussion group.

  • Also from the arxiv, nice article by Nick Barton and colleagues on the role statistical mechanics has played in evolutionary biology.

  • Interesting work on the arxiv on critical transitions from Christian, a post-doc in Thilo Gross’s group at Max Planck Complex Systems, Dresden.  Also meaning to look over there generalized models in food webs paper (Gross et. al. 2009).

  • Looks like a fantastic article on estimating diversification (speciation/extinction) rates in PNAS(Stadler, 2011), and an accompanying package, TreePar (which actually looks like it’s been on CRAN since last June).

  • Looking over this paper led me to a few others also by Tanja, discussing some of the challenges of simulating(Stadler, 2011) and sampling trees(Hartmann et. al. 2010).

  • See TreePar discussion in later post.

References