Edgar Dobriban

Edgar Dobriban
  • Assistant Professor of Statistics

Contact Information

  • office Address:

    455 Jon M. Huntsman Hall
    3730 Walnut Street
    Philadelphia, PA 19104

Research Interests: High-dimensional asymptotics, random matrix theory, multiple testing.

Overview

My research focuses on statistical methods for “big” data. On the theoretical side, I leverage results from random matrix theory for the analysis of multivariate data when the dimension and sample size are large. On the applied side, I have developed methods for multiple testing motivated by the genomics of exceptional human longevity.

I obtained my PhD in Statistics from Stanford University in June 2017. I’ve had David Donoho as my PhD advisor and collaborated with Art Owen, with Stuart Kim’s lab, and with Amit Singer’s group. In 2012, I obtained a BA in mathematics from Princeton University.

Continue Reading

Research

Talk slides: GitHub.

Teaching

Current Courses

  • STAT430 - Probability

    Discrete and continuous sample spaces and probability; random variables, distributions, independence; expectation and generating functions; Markov chains and recurrence theory.

    STAT430401 ( Syllabus )

    STAT430402 ( Syllabus )

  • STAT510 - Probability

    Elements of matrix algebra. Discrete and continuous random variables and their distributions. Moments and moment generating functions. Joint distributions. Functions and transformations of random variables. Law of large numbers and the central limit theorem. Point estimation: sufficiency, maximum likelihood, minimum variance. Confidence intervals.

    STAT510401

    STAT510402

  • STAT991 - Sem In Adv Appl Of Stat

    This seminar will be taken by doctoral candidates after the completion of most of their coursework. Topics vary from year to year and are chosen from advance probability, statistical inference, robust methods, and decision theory with principal emphasis on applications.

    STAT991303

Past Courses

  • STAT430 - PROBABILITY

    Discrete and continuous sample spaces and probability; random variables, distributions, independence; expectation and generating functions; Markov chains and recurrence theory.

  • STAT510 - PROBABILITY

    Elements of matrix algebra. Discrete and continuous random variables and their distributions. Moments and moment generating functions. Joint distributions. Functions and transformations of random variables. Law of large numbers and the central limit theorem. Point estimation: sufficiency, maximum likelihood, minimum variance. Confidence intervals.

  • STAT991 - SEM IN ADV APPL OF STAT

    This seminar will be taken by doctoral candidates after the completion of most of their coursework. Topics vary from year to year and are chosen from advance probability, statistical inference, robust methods, and decision theory with principal emphasis on applications.

Data

Software

This page has links to software implementing methods developed in my papers. The software is usually hosted on  GitHub. That page also contains software to reproduce the computational results of my publications.
Feel free to contact me if you are interested in using this software.

ePCA: github Matlab

Contains the ePCA method for principal component analysis of exponential family data, e.g. Poisson-modeled count data.
Implements methods for denoising individual datapoints. (with L.T. Liu)

Related paper: Liu et al., 2016;

EigenEdge: github Matlab

Contains methods for working with large random matrices, including

  • The SPECTRODE method for computing eigenvalue distributions of covariance matrices (general Marchenko-Pastur distributions).
  • Methods to compute moments and quantiles of these distributions.
  • Optimal linear spectral statistics for testing in principal component analysis.
  • Tools for spiked covariance models: spike and cosine descriptors, optimal shrinkers.

Related papers: Dobriban, 2015; Dobriban & Wager, 2015; Dobriban, 2016;

pweight/Princessp: CRAN. github R. github Matlab

Implements P-value weighting techniques for multiple hypothesis testing. These methods can improve power in multiple testing, if there is prior information about the individual effect sizes. Includes the iGWAS method designed for Genome-Wide Association Studies.

Related papers: Dobriban et al., 2015; Fortney et al., 2015; Dobriban, 2016;