Peking: Personal statement

When I plotted the malaria incidence data my uncle had collected as a county health officer in Yunnan, the curve refused to behave. ¹He expected cases to fall smoothly after the bed-net program; instead they dropped, plateaued, then crept back up. He blamed the weather. I suspected the model. That argument, carried out over instant noodles in his clinic, is the reason I now want to study statistics, and specifically the statistics of populations, at Peking University. I spent the following summer teaching myself enough R to test both of us. I learned that his "plateau" was an artifact of how the township aggregated reports, and that once I disaggregated by village the resurgence was real, concentrated in three settlements near a new reservoir. ²I was wrong about the cause and right about the method, and the experience taught me something I have not been able to put down since: that the difference between a number that misleads and a number that saves a life is usually a question of how you model the noise. To prepare, I went looking for the harder version of that question. In school I exhausted our mathematics offerings by the eleventh grade, so I worked through Casella and Berger on my own, stalling badly at sufficient statistics until a retired professor in our city agreed to meet me on weekends. ³He did not simplify the material; he made me re-derive it until I could explain why the factorization theorem had to be true and not merely that it was. I failed his first three attempts at a likelihood-ratio proof and got the fourth. That ratio of failure to success is, I have come to think, an honest preview of research, and I would rather know now that I can sit inside a problem that refuses me for weeks than discover it too late. My reason for Peking specifically is not that it is in China; I am already in China. ⁴It is the work coming out of the School of Mathematical Sciences and the Center for Statistical Science, where the genetics-and-statistics overlap I care about is treated as one subject rather than two departments lending each other tools. I read a recent paper from the Center on controlling false discovery in genome-wide association studies and understood maybe half of it, which is exactly the gap I want to be closed by sitting in those seminars rather than reading them alone in Yunnan. ⁵I want supervisors who will tell me, as my weekend professor did, that my proof is not yet a proof. I am aware of what I am missing. My measure theory is informal, my coding is self-taught and probably full of habits a real statistician would wince at, and I have never worked with a dataset I did not, in some sense, already understand. ⁶Peking is where I expect those comforts to be taken away from me usefully. My uncle still emails me his county's numbers every month, and the resurgence near the reservoir has not been explained. ⁷I would like to learn enough, in the right place and from the right people, to one day write him back an answer that holds up.