Computational statistics postgraduate personal statement example

Personal statement example

My degree gave me the theory of statistical inference; what drew me in was the point where the theory runs out and a computer has to take over. That point arrived for me in a second-year lab exercise on confidence intervals for a median. The textbook formula depended on assumptions my data plainly did not meet, and the recommended fix was to resample the data several thousand times. I found the idea startling at first, then persuasive, and it shaped the rest of my studies.

I chose a final-year project comparing bootstrap and Bayesian approaches to quantifying uncertainty with small samples. Working in R, I implemented the percentile and BCa bootstrap, then a simple Metropolis-Hastings sampler for the same problem, and spent longer than I expected on the unglamorous parts: choosing proposal variances, checking trace plots and effective sample sizes, and working out why my chains agreed with each other but not with my analytic answer. The bug was mine, in the log-likelihood, and finding it taught me more about the method than any derivation had. I also learned to be careful about what computation can and cannot rescue: resampling twelve observations gave me honest arithmetic about a very uninformative dataset. My supervisor's suggestion that I read Efron and Hastie's Computer Age Statistical Inference was well timed, and the book's framing of algorithms and inference as two intertwined strands is close to how I now think about the subject. Alongside it I worked through parts of James, Witten, Hastie and Tibshirani's An Introduction to Statistical Learning, mainly to understand cross-validation properly rather than as a button to press.

My modules in probability, linear algebra and numerical analysis are the background I want to build on, and I am aware of the gaps. I am comfortable with Markov chains and matrix decompositions but have had little exposure to Monte Carlo methods beyond the basics, to computational approaches for high-dimensional models, or to writing code that others have to read. Since graduating I have been rebuilding my project as a small R package with proper documentation and unit tests, which has been a useful corrective to the habit of writing scripts that only work once.

Outside university, I work part-time as a library assistant and tutor GCSE and A-level maths, where explaining conditional probability to sceptical sixteen-year-olds has improved my own clarity. Last year my aunt, who runs a small upholstery business, asked me to help her make sense of three years of sales records kept across several spreadsheets. Most of the work was tedious data cleaning, and the most useful output was a simple seasonal summary rather than anything sophisticated; it was a good lesson in matching the method to the question and the data actually available. I also play contract bridge at a local club, which has given me an intuitive feel for reasoning under uncertainty and for updating beliefs on partial information — and regular evidence that my intuition benefits from being checked against calculation.

A master's in computational statistics is the natural next step for me because I want formal training in the methods I have only glimpsed: MCMC and its diagnostics, resampling, optimisation, and simulation-based inference, together with the software practice that makes such work reproducible. My aim afterwards is to work as an applied statistician, ideally in health or public-sector research, where careful estimates of uncertainty matter to the decisions people make. I would be glad of the chance to prepare for that properly.