Student SPHere 2008-09

So much has happened this past couple of weeks, and so much is going to happen in the next month or so, it can feel almost overwhelming at times. My biggest “accomplishment” was turning in a completed first draft of my Plan B to my advisor this week. I really have learned a lot from my project, and many of the lessons were hard ones to learn. I am not the most “syntax savy” individual, so using an “unpopular” software for my analyses was probably not the most prudent move on my part. However, the software—Mplus—is the “best” software available for doing analyses with latent variables methods, which was the topic I wanted to explore in my Plan B, so use it I did. Unlike more common packages like SAS and R, and even SPSS, there are only a scant number of recources one can utilize with Mplus questions and issues. The most painful part was having to scour the internet for HOURS at a time to glean little bits of syntax buried deep within someone’s rambling webpage that would enable my program to even run. Since my data set is rather sparse for use in a latent variables approach, and a lot of my models were rather complex, I ran into almost every issue possible when attempting to run my models. From failure to converge, to model nonidentification, to non positive-definitive matrices to unstable variance estimates and Lord only knows what else, Mplus fought me every step of the way. Of course, the difficulties I encountered have contributed to a pretty interesting discussion session where I address these issues in terms of the tradeoff between Mplus’ flexibility of estimation and the copious issues that can accompany its utilization for analysis.
I also did manage to finish my Bayes take home exam. Now, when I say “finish,” I simply mean that there was something written down for every question; however, I am more than convinced that I fell far short of the goal of actually correctly answering most of the problems. I am proud of the fact that I did finish more than I thought I would, and that my WinBugs and R code at least produced answers that seemed at least related on some level to the test questions. If this were not my fun class that I’m taking pass/fail, I would be much more stressed than I currently am, but with things the way they currently are, I am happy with how the test went. The Bayes class itself continues to rock my socks off. We are now studying spatial Bayes, in which correlations among observations based on spatial proximity (such as counties in a state that share a border) are taken into account. We learned how to import maps into R and Winbugs and then analyze data for those maps along with map-specific covariates (such as average sun exposure by county) and examine how the relative observed responses can change with the adjustment for these covariates. As a toy example from class, the raw average SAT scores tend to be higher in the Midwest; however, when the number of people taking the SAT is taken into account, the higher scoring areas appear on the coasts. This is probably due to the fact that almost everyone on the coasts take the SAT—not the ACT—and so you have students of all abilities reporting SAT scores, whereas in the Midwest, most students take the ACT, and only those students who are of a high enough academic caliber to be applying to the better schools on the coasts will bother to take the SAT, thus only “good” students who will tend to score high are taking the test in these areas.
In Clinical Trials, I am just trying to finish up the slides for our big group presentation at the beginning of May. Part of the requirements for the class is to complete a fake (and abbreviated) study protocol as a group and then present the protocol to the class. I am one of the presenters, and as such it is part of my job to make sure that our presentation is adequately cogent. Luckily, my group has been able to work well together, and I feel pretty good about the preparation levels for the end of this class. This is also the only class where I will actually have to take a final, so one my presentation is over, I will be able to devote some time to studying for that.
I have two other presentations for which to prepare—one for my Translational Research class on cost effectiveness studies, and one for my Bayes class. I have finished the former and will be (hopefully) completing the latter very soon (as long as Plan B revisions don’t take over my life). There is, of course, the Plan B presentation, which I will have to start preparing very soon as well.
This week I’ll also be preparing to go to New York for a cousin’s wedding this Friday. While it is not the most convenient time for me to be taking a trip, I will be really glad to see my extended family and unwind a little bit. It will also be really nice to get back to the City and meet up with some old friends from college. I don’t get to see them (or New York) as often as I would like now that I live in Minnesota, but every time I get to go back, I always have a great time.