Department of Mathematics and Statistics, Boston University  
line decor
  HOME  ::  
line decor


My general research interest is in the statistical modeling of instrumental data in temporal, spatial, and network-indexed contexts. Most of the problems on which I currently work involve the statistical analysis of network data, typically motivated by collaborations with colleagues in bioinformatics, computer science, geography, and signal processing. Much of my earlier work, and some of my current work, has focused on the statistical modeling of scale, motivated by problems in astronomy, epidemiology, geography, and signal and image processing.

Current research includes work on the following projects.

  1. Statistical Foundations for Measurement-based System Verification in Complex Networks
    (Supported by AFOSR award 12RSL042.)

    Focus: We pursue a broad-based, multi-faceted program of research to systematically lay key pieces of the statistical foundation necessary to pursue measurement-based systems validation in complex networks. Work includes efforts in network sampling, confidence intervals for network summary statistics, inference for network model parameters, multiscale network denoising, and multi-attribute network analysis.

  2. CRCNS: Dynamic Network Analysis of Human Seizures for Therapeutic Intervention
    (Supported by NIH award 1R01NS095369-01.)

    Focus: Develop and apply network-based tools for analysis of human seizure data during onset and termination, towards the goal of suggesting possible therapeutic interventions.

    Collaborators: Mark Kramer (Math/Stat, Boston University); Sydney Cash and Catherine Chu (MGH).

  3. Statistical Foundations for Analysing Large Collections of Network-Data.
    (Supported by ARO award W911NF-15-1-0440.)

    Focus: Development of analogues of "Statistics 101" tools for data sets consisting of network data objects, through a combination of principles and techniques from geometry, probability, and statistics.

    Collaborators: Steve Rosenberg (Math/Stat, Boston University) and Lizhen Lin (Statistics and Data Sciences, UT-Austin).

Previous research has included, among other things, work on the following projects.

  1. Wide-Aperture Traffic Analysis for Internet Security
    (Supported by NSF grant CNS-0905565.)

    Focus: Development of methods for identifying malicious and unwanted Internet activity, particularly low-volume activity "hidden" among normal activity.

    Collaborators: Mark Crovella (Computer Science, Boston University) and Paul Barford (CS, University of Wisconsin).

  2. Multi-cohort, Network-guided Regression for GE/GG Interactions in Disease Traits
    (Supported by NIH award 1R21ES020827.)

    Focus: Development a principled and biologically-informed two-stage network-guided statistical methodology to detecting GxE and GxG interactions associated with human disease in current large-scale, multi-cohort association analyses. Ultimately, our work should help to significantly acce lerate the development of targeted therapies and personalized medicine strategies, through its fundamental impact on the early stages of the overall process.

    Collaborators: Josee Dupuis (Biostatistics, Boston University).

  3. Predicting Drug Mechanism via Chemo-Genomic Profiling and Sparse Simultaneous Equation Models of Gene Regulation
    (Supported by NIH award GM078987.)

    Focus: Development of statistical and computational methods for predicting the mechanism of action of a proposed drug using gene expression profiles of drug activity. Methodology is grounded on the use of simultaneous equation models and complexity penalized inference procedures, with the aim to exploit the expected sparseness of these models.

    Collaborators: Scott Schaus (Chemistry, Boston University).

  4. Statistical Propagation of Low-Level Uncertainty to High-level Knowledge and Decision-Making in Network Information Environments
    (Supported by ONR award N000140910654.)

    Focus: Creation of a methodological foundation, with accompanying theoretical and computational components, for propagating uncertainty from `low-level' data sources to `high-level' network-based knowledge tasks.