I am interested in developing robust applicable statistical procedures for high-dimensional data, with rigorous theoretical guarantees, under minimal assumptions. My graduate research focuses on the study of likelihood based inference in high-dimensional generalized linear models. In particular, it uncovers that for a class of generalized linear models, classical maximum likelihood theory fails to hold in a high-dimensional setup. Consequently, p-values/confidence intervals obtained from standard statistical software packages are often unreliable. Some illustrations to this end can be found here.
In a series of papers, I have studied several aspects of this phenomenon and have developed a modern maximum-likelihood theory suitable for high-dimensional data that accurately characterizes properties of likelihood based approaches and that can be used by practitioners to obtain valid inference for such non-linear models.
I also have a standing interest in controlled variable selection and possible connections to causal inference. I am simultaneously involved in research on the different definitional aspects of algorithmic fairness, their connections, limitations and implementations, and robust metric learning for fairness.
P. Sur, Y. Chen, and E. J. Candès. The likelihood ratio test in high-dimensional logistic regression is asymptotically a rescaled chi-square. Probability Theory and Related Fields, to appear, 2018+. [pdf] [supp] [talk] [arXiv]
P. Sur, G. Shmueli, S. Bose, and P. Dubey. Modeling bimodal discrete data using Conway- Maxwell-Poisson mixture models. Journal of Business and Economic Statistics, Volume 33, 2015 - Issue 3. [pdf] [journal]
S. Bose, G. Shmueli, P. Sur, and P. Dubey. Fitting COM-Poisson mixtures to bimodal count data. Proceedings of the 2013 International Conference on Information, Operations Management and Statistics (ICIOMS 2013), winner of Best Paper Award. [conference]