Student SPHere 2008-09

Last weekend, I fully embraced the spirit of autumn. On Saturday, I baked cupcakes using selections from fall's delicious bounty: beets, carrots, maple syrup, and chevre. Ok, maybe that last one's not particularly seasonal. They came out so well! My sweet tooth is strong, but my love of complex flavors is even stronger. These were earthy, rich, tangy, and sweet all at once-- yum! The recipe can be found at one of my favorite food blogs, Apartment Therapy: The Kitchn. I made a full recipe of the cake and a half recipe of the frosting (altered only to incorporate some of the vanilla beans I bought recently) and baked them in cupcake tins instead of assembling as a layer cake.

Sunday, I visited the MN Landscape Arboretum for the first time. What a gorgeous place! It seemed that approximately everyone else in the state had the same idea-- it was crowded with wedding parties, families taking portraits, and couples enjoying the crisp air. The ground are expansive, so we were able to wander off onto quiet wooded paths. This is truly a gem of Minnesota; I only wish I hadn't waited three years to visit. Next time, I will make sure to plan time to visit the apple house for some U of M developed apple varieties. P.S. Students at the U of M always get free admission!

This week has been the usual array of challenges and opportunities at school. My former colleagues in the Division of Epidemiology have kicked my butt into gear to re-submit a couple of manuscripts that have languished since I left last year. Thanks, guys! So I hope that soon I will have a couple more pubs to add to the old CV. Speaking of which, I updated it as part of the application process to the PhD program in Biostats. I'm sure that SOPHAS is a great tool for new students applying to a number of schools. But having used it twice now to apply to just one school (the U of M, where I was already a student both times), it's starting to get on my nerves. None of the information is saved from previous application cycles, so I had to repeat the hours-long process of manually entering every class I've ever taken. Whew-- I've been taking college coursework for 10 years now. Seriously, when people joke about being a professional student, I'm pretty sure I'm exactly what they have in mind.

In Probability and Linear Models, we have embarked on the core material, having finished mathematical preliminaries. It's pretty interesting to learn what happens inside the black box of SAS or R when one fits a generalized linear mixed model, for example. One huge benefit of computing has been to find ways around the hurdle of high-dimensional integrations. This is a concept that keeps coming up again and again. Statistical estimation relies on the likelihood, which is generally a product over observations. If we want to rid the likelihood of some parameters (like random effects in the case of mixed models or the model parameters themselves in the case of obtaining a marginal for Bayesian inference) the necessary integration can get hairy very quickly. With modern computing, we can take a variety of approaches: Expectation-Maximization (EM) algorithm, Monte Carlo integration, quasi-likelihood optimization, Gaussian quadrature, etc. etc. Good stuff!