|Eric V. Slud
Professor, Statistics Program
Department of Mathematics
University of Maryland
College Park, MD 20742
Info on Older RIT's
Office hours: W 1:30-3pm, or by appointment (MWF only)
Mini-Course on Cross-Classified Factor
(10/17/05) Mathematical Challenges in Cross-Classified
Summary of interesting mathematical issues related to PhD theses about Factor Analysis
by my former students Yang Cheng(2004) and Sophie (Hsiao-Hui) Tsou (2005).
Mini-Course on Markov Chain Monte Carlo
(Statistical Simulation Techniques) Spring '04: 4/21, 4/28
Slides can be found at links indicated under each lecture, covering the following topics:
1 (4/21/04) Metropolis Hastings Algorithm --
Motivation from Accept-Reject
simulation methodology and from Markov Chain theory. Extended example and issues
involved in the choice of `proposal Markov chain' from which the Metropolis-Hastings
chain is built. Gibbs Sampler motivation.
Lecture 2 (4/28) Recap of Gibbs Sampler motivation. Testing for Markov Chain Monte Carlo
convergence from the internal evidence of the Gibbs Sampler trajectory. Statistical examples:
Bayesian statistical computation and frequentist treatment of hierarchical statistical models.
Mini-Course on Statistics of Survival Data
(Fall '02: 11/6, 11/13, 11/20)
Slides can be found at links indicated under each lecture below.
The lecture topics are as follows:
(11/6/02) Survival Times, Death Hazards & Competing Risks
Lecture 2 (11/13) Population Cohorts and Martingales
Lecture 3 (11/20) Survival-data likelihoods with Infinite-Dimensional Parameters
(1) Fall '05 RIT on
Statistics of Models with Growing Parameter Dimension
(2) Spring '04 RIT on Meta-Analysis. Click
Briefly, meta-analysis concerns the simultaneous statistical analysis of a number of
related studies or datasets within a single statistical model. The fact that parameters are
shared across datasets (e.g. a treatment-effectiveness parameter assumed constant
across a number of separately conducted clinical studies of the effectiveness of the same
treatment regimen for the same disease) allows the possiblity of increasing sensitivity or
power of statistical tests. However, such an increase in precision comes at the price of
simultaneous model assumptions whose compatibility with the data must be validated.
This RIT was an outgrowth of the Fall '03 RIT on Large Cross-Classified Datasets
(see web-page linked below for details).
(3) Spring and Fall '03 RIT on Statistics of Large Cross-Classified Datasets:
see RITF03 web-page .
(4) Intensive Seminar, Fall 2002. See plan for details.
In Fall 2002, I ran a `research interaction' seminar including my own graduate advisees
and others, on the mathematical & statistical topics which more broadly correspond to
the overlap of my students' thesis projects and most of my own current research interests,
namely Statistics of Large Cross-Classified Datasets. Roughly speaking, these are
problems in which there is a large sample-size n, but where the predictor variables
and/or cross-classifications of the sample units become more complicated or numerous
as n gets large. Such problems range from Semiparametric Statistical Inference to
Order-selection problems in regression and time series, to Classification and Clustering
as in the Microarray data problems mentioned below. These problems suggest the
need for a new Asymptotics which explicitly recognizes the growth of the parameter-
space of a probability model as a function of the size n of the dataset.
(5) Intensive Seminar, Spring 2002. See plan for details.
In Spring 2002, following up on the Fall 2001 seminar described below, I ran an intensive
seminar on statistical analysis of DNA Microrarrays, for students considering research
in this area. Data-analysis figured prominently, performed by me and also by two of the
several graduate students who participated.
(6) Genomics/Microarray Seminar Fall 2001, AMSC 699:
Mathematical Topics in Functional Genomics. Click here for the reading list.
Other Past Teaching and Seminars
Spring '04, introductory course Stat 470
on Actuarial Mathematics, taught primarily
Fall '03, Stat 798S, topics course on Survival
Analysis . For slides of my Stat Seminar
Talks on various topics are preserved in the directory OldTalks and briefly described and linked below:
(I) Census statistics,
specifically demographic modelling of nonresponse to national surveys, with particular application to Weighting Adjustment and Small Area Estimation (SAE).
Much of my small-area estimation work has been directed toward the SAIPE (Small Area Income and Poverty Estimation ) program of the
Census Bureau. See for example a Small
Area Estimation model-comparison study about SAIPE that I wrote. My methodological research in this area includes
small-area and MSE estimation from survey data satisfying nonlinearly transformed Fay-Herriot models
or left-censored Fay-Herriot models.
My Discussion of a Review Paper of JNK Rao on Small Area Estimation mentions several research directions in this problem area that are still highly relevant.
Some further work on internal evaluation of biases due to weighting adjustment for nonresponse in a longitudinal survey (SIPP, Survey on Income and Program Participation) is described in my Nov. 2007 FCSM talk. A paper describing the contents of that talk more fully can be found here, and in a form that appeared in the Journal of Official Statistics, here. Other recent work on simultaneous nonresponse-adjustment and calibration of weights in complex surveys can be found in a Census SRD Technical Report.
Miscellaneous other projects related to Sample Survey design and estimation, with particular reference to Census Bureau problems, have been the topics of presentations I have given over the past 15 years at Joint Statistical Meetings. Many of these can be found in my Census statistics web-directory.
(II) Survival data analysis, which includes both semiparametric inference and clinical trial design issues, as well as a selection of my journal papers on biostatistical survival analysis. The semiparametric work emphasizes maximization of variants of nonparametric likelihoods, especially in Transformation and Frailty models. A general approach to efficient semiparametric estimation described in slides from a talk given in the IISA Conference, June 14, 2002. Other work relates to decision-theoretic optimal early-stopping procedures and new designs in clinical trials.
For slides of a Stat Seminar I gave in Fall '03 at NIH on asymptotic theory of Semiparametric statistical procedures in Transformation models, click here.
A 2014 paper I wrote with a student, Jiraphan Suntornchost, describes models arising in survival analysis, but also in reliability and other fields, allowing development of flexible families of parametric densities for survival times. For a Stat Seminar talk I gave on that research, see Seminar on parametric survival densities. A talk on Bivariate Marker-Degradation Processes in Reliability was based on the thesis of my student Vasilis Sotiris and presented at the Mathematical Methods in Reliability Conference in Beijing in 2011.
(III) Meta-analysis, in biostatistics: I wrote two papers on this topic, in 2011 and 2018.
(IV) Pharmaceutical Statistical Methods: with my student Meiyu Shen (of FDA, co-advised with Estelle Russek-Cohen also of FDA), I have worked on several aspects of the design of regulatory clinical trials, in particular on sample size calculation for two-period crossover bioequivalence trials and on distribution of noisy measurements generated from pharmacokinetic models.
(V) Large-scale data problems with emphasis on cross-classified data, Principal Components (paper on representation of tongue surface during speech, appeared in the journal Phonetica), and clustering. More recently, I have had two students (Yang Cheng and Sophie Tsou) obtain PhD's working on Factor Analysis models. A talk I gave on this work in 2005 [and then again in the Diffusion Wavelet RIT in Fall 2007] can be found here.
(VI) Stochastic processes. Two examples are work emphasizing high-dimensional Markov processes applied to equilibria in Economics (paper in Journal of Economic Theory, for which 2nd pdf file in directory contains Figure); to Protein-folding; and to ascertainment of number of distinct DNA `species' from sequencing experiments.