How it works:
Please choose a paper from the following list.
Look at the paper first and let
me know if you would like to work on that paper so that
I can mark it as unavaliable on this website.
Papers that are
overstriken are no longer
Contact me (in person or by e-mail) to
schedule a meeting so that we can discuss exactly
what you would like to do with the paper you chose.
- Athey, S., Tibshirani, J. and Wager, S. (2019). Generalized
random forests. The Annals of Statistics, 47(2), 1148 - 1178.
This paper proposes a new way of computing local estimators, where
random forests are used to identify "neighbour" points in the training
set, instead of the
usual kernel-based concept of "local". This approach appears to neatly circumvent
the curse of dimensionality, and can in principle be applied to a large
class of estimators.
Fan, J., Li, Q. and Wang, Y. (2017). Estimation of high dimensional
mean regression in the absence of symmetry and light tail assumptions. Journal of the
Royal Statistical Society. Series B, Statistical methodology, 79-247.
This paper shows that, in theory, one can consistently
and robustly estimate
a regression function when the errors have an asymmetric distribution.
However, the devil is in the methodological and practical details.
If you pick this paper, you will focus on running numerical
the proposed method
in realistic settings.
- Fasiolo, M., Pya, N. and Wood, S.N. (2016). A comparison of inferential methods for
highly nonlinear state space models in ecology and epidemiology. Statistical Science,
31(1), 96-118. DOI: 10.1214/15-STS534
Wood, S. (2010). Statistical inference for noisy nonlinear ecological dynamic systems.
Nature, 466(26), 1102-1104, DOI: 10.1038/nature09319
These two papers should be read together. They argue that the
usual way to perform inference for dynamic systems when they
are "almost chaotic" may intrinsically be unreliable, and
propose synthetic likelihood as a feasible alternative.
- Kantas, N., Doucet, A., Singh, S.S., Maciejowski, J.
and Chopin, N. (2015). On particle methods for parameter
estimation in state-space models. Statistical Inference,
30:3, 328-351. DOI: 10.1214/14-STS511
This review paper discusses
parameter estimation (either MLE or Bayesian)
for non-linear state-space models (aka Hidden Markov models)
Monte Carlo (particle methods). These algorithms are
very useful when working with complex models (e.g.
models using phylogenetic and transmission trees).