Statistics Seminar Series

Department of Mathematical Sciences
and
Center for Applied Mathematics and Statistics

New Jersey Institute of Technology


Fall 2009

 

All seminars are 4:00 - 5:00 p.m., in Cullimore Hall Room 611 (Math Conference Room) unless noted otherwise. Refreshments are usually served at 3:30 p.m., and talks start at 4:00 p.m. If you have any questions about a particular seminar, please contact the person hosting the speaker.

 

Date
Speaker and Title
Host

Wednesday
September 9, 2009
4:00PM

Sheldon Ross, University of Southern California
Multiple Item Selling Problems (abstract)

Manish Bhattacharjee

Thursday
September 17, 2009
4:00PM

Glen Atlas, University of Medicine and Dentistry of New Jersey, Newark, NJ
and Stevens Institute of Technology, Hoboken, NJ
Development of a Recursive Finite Difference Pharmacokinetic Model from an Exponential Model: Application to a Propofol Infusion (abstract)
Sunil Dhar

Thursday
September 24, 2009
4:00PM

Li Wang, Ph.D., Global Biometric Sciences-Bristol-Myers Squibb Company
Orthogonal blocking of response surface split-plot designs (abstract)
Sunil Dhar

Thursday
October 15, 2009
4:00PM

Das Purkayastha, Ph.D., Novartis Pharmaceuticals
A Methodological Perspective of Predicting Circadian Fluctuations of 24-Hour Ambulatory Blood Pressure: A New Look to ABPM Analyses in Cardiovascular Clinical Trials (abstract)
Sunil Dhar

Thursday
October 29, 2009
4:00PM in
Cullimore Lecture Hall I

Abhijit Dasgupta, Ph.D., CEO ARAAstat
Exploration of High Dimensional Data Using a Flexible Learning Method (abstract)
Sunil Dhar
Thursday
 November 5, 2009
 4:00PM in
Cullimore Lecture Hall I
Yi-Hsuan Lee, Ph.D., Educational Testing Service
Controlling Item Exposure in Multidimensional Computerized Adaptive Testing (abstract)
Chung Chang
Thursday
 November 12, 2009
4:00PM
Dr. Zhi Wei, Department of Computer Science, NJIT
An HMM-based Optimal Multiple Testing Procedure for Genome-wide Association Studies (abstract)
Wenge Guo
Thursday
 November 19, 2009
4:00PM
Jon Kettenring, Drew University
Massive Datasets (abstract)
Ari Jain
Thursday
 December 3, 2009
Cullimore Lecture Hall I
4:00PM
Md. Aleemuddin Siddiqi, Ph.D., Symbiance, Inc., Princeton Junction, NJ
Analysis of Microtubule Dynamics Using Growth Curve Models (abstract)
Sunil Dhar

 

 

 

 

ABSTRACTS

Multiple Item Selling Problems:

We consider models in which there are n identical items to sell. Independent and identically distributed offers arrive sequentially and must either be accepted or rejected. The initial model assumes that a single individual is selling all n items. Assuming a cost per time period per unsold item, his objective is to maximize his expected return. We then suppose that there are n sellers, each wanting to sell a single item, who are lined up in a fixed order. An offer initially goes to the seller first in line; if rejected it goes to the next in line, and so on. Upon accepting an offer the seller departs. With all sellers assuming correctly that everyone wants to maximize their expected return, does the expected total return equal that of the equivalent one-seller model? What if the seller ordering is not fixed but randomly changes before each offer?

We also consider a model, similar to the initial one, but which now supposes that each of the n buyers pays not what they offered but the minimal of the accepted offers.

Sheldon Ross ~ September 9, 2009

Development of a Recursive Finite Difference Pharmacokinetic Model from an Exponential Model: Application to a Propofol Infusion:

Pharmacokinetic models, using recursive finite difference equations (RFDEs), can be derived directly from traditional exponential models. This method has been successfully applied to propofol infusion data. Furthermore, this technique yields identical accuracy, on a subject-specific basis, as the exponential model from which each RFDE model was derived. Specifically, these infusion models are based upon an inhomogenous RFDE: P(k+3) = A·P(k+2) + B·P(k+1) + C·P(k) + R, where A, B, C, and R are non-zero constants and P represents plasma propofol levels for each kth unit of time. When applied to propofol infusions, RFDE modeling has advantages, over traditional exponential models, in that fewer coefficients are needed and patient-to-patient variation of these coefficients is reduced. However, initial conditions for RFDEs have to be specified. These characteristics, of RFDE modeling of propofol infusions, are similar to those for RFDE modeling of propofol boluses. Based on these findings, as well as those of our prior study, RFDE pharmacokinetic modeling can be applied to both infusion and bolus data of propofol. Further research, on the applications of RFDEs in pharmacokinetics, appears warranted.

Glen Atlas, Department of Anesthesiology, University of Medicine and Dentistry of New Jersey, Newark, NJ and the Department of Chemistry, Chemical Biology and Biomedical Engineering, Stevens Institute of Technology, Hoboken, NJ ~  September 17, 2009

Orthogonal blocking of response surface split-plot designs:

When all experimental runs cannot be performed under homogeneous conditions, blocking can be used to increase the power for testing the treatment effects. Orthogonal blocking provides the same estimator of the polynomial effects as the one that would be obtained by ignoring the blocks. In many real-life design scenarios, there is at least one factor that is hard to change, leading to a split-plot structure. This paper shows that for a balanced ordinary least square-generalized least square equivalent split-plot design, orthogonal blocking can be achieved. Orthogonally blocked split-plot central composite designs are constructed and a catalog is provided.

Li Wang, Ph.D, Sr. Research Biostatistician, Biostatistics and Programming ~ September 24, 2009

A Methodological Perspective of Predicting Circadian Fluctuations of 24-Hour Ambulatory Blood Pressure: A New Look to ABPM Analyses in Cardiovascular Clinical Trials:

Ambulatory blood pressure monitoring (ABPM) data are generally nonstationary and stochastic like many other clinical parameters. Clinical observations of 24-hour ambulatory data indicate such variations of blood pressures. But conventional analyses of ABPM data fail to capture such variability. This paper provides a new approach of analyzing ambulatory blood pressure by smoothing local peak and trough of blood pressure over time. It is an iterative modal decomposition method that uses a monotonic intrinsic mode function. Also, using a family of exponential Gumbel-extreme value distributions it has been observed that the method maintains its predictive power unlike other traditional approaches. This new computationally feasible approach of analyzing ABPM data is tested using real-life clinical data for prediction of peak and trough blood pressure during the morning, day, and night time.

Das Purkayastha, Ph.D., Director, Biostatistics - Biometrics, U.S. Clinical Development ~ October 15, 2009

Exploration of High Dimensional Data Using a Flexible Learning Method:

High dimensional data is now widely prevalent, especially in areas related to genomics and proteomics in the biological and medical literature. A major focus in the statistical literature related to high dimensional data has been on testing hypotheses and the accompanying multiple comparisons issues. However, in many experiments, there are not sufficient resources to achieve the sample sizes necessary to make testing possible with reasonable power. The p-values generated by hypothesis testing in this context then merely serve as a ranking procedure to inform subsequent studies. The current work proposes an exploratory method for investigating high dimensional data and extracting information without taking a hypothesis testing approach. The QR matrix decomposition will be used to provide informed dimension reduction of the data, which can then be utilized for clustering, classification and exploration of the data.

Abhijit Dasgupta, Ph.D., CEO ARAAstat ~ October 29, 2009

Controlling Item Exposure in Multidimensional Computerized Adaptive Testing:

Although computerized adaptive tests have enjoyed tremendous growth, solutions for important problems remain unavailable. One problem is the control of item exposure rate. Because adaptive algorithms are designed to select optimal items, they choose items with high discriminating power. Thus, these items are selected more often than others, leading to both overexposure and underutilization of some parts of the item pool. Overused items are often compromised, creating a security problem that could threaten the validity of a test. We propose a strategy based on stratification in accordance with a functional of the vector of the discrimination parameter, which can be implemented with minimal computational overhead. Both theoretical and empirical validation studies will be discussed. Empirical results indicate significant improvement over the commonly used method of controlling exposure rate that requires only a reasonable sacrifice in efficiency.

Yi-Hsuan Lee, Ph.D., Associate Research Scientist, Educational Testing Service ~ November 5, 2009

An HMM-based Optimal Multiple Testing Procedure for Genome-wide Association Studies:

Genome wide association studies (GWAS) interrogate common genetic variation across the entire human genome in an unbiased manner and hold promise in identifying genetic markers with moderate or weak effect sizes. However, most conventional testing procedures ignore dependency among markers and suffer from severe loss of efficiency in GWAS. In this talk, I will present a data-driven testing procedure (PLIS), which employs hidden Markov Models to exploit dependency information among adjacent SNPs. PLIS is shown to control the false discovery rate (FDR) at the nominal level while have the smallest false negative rate (FNR) among all valid FDR procedures. By re-ranking significance for all SNPs with dependency considered, PLIS gains higher power than conventional p-value based methods. Simulation results and the application to a GWAS T1D dataset demonstrate that our proposed approach has better reproducibility and yields more accurate results. Some extensions will be discussed at the end.

Dr. Zhi Wei, Assistant Professor, Department of Computer Science, NJIT ~ November 12, 2009

Massive Datasets:

Massive datasets are so labeled because of their size and complexity. They do not yield readily to standard statistical analyses. The resulting frustration has served as a spur to researchers to develop better tools. Some progress has been made, but the need for considerably more explains why this line of research remains a top priority. Interdisciplinary teamwork is at least as important as tools and can be the key to cracking the hard challenges that these datasets pose. This overview talk includes background information, examples, and statistical strategies to illustrate the state of the art. (Reference: Wiley Interdisciplinary Reviews: Computational Statistics, 2009, 25-32.)

Jon Kettenring, Statistics Professor, Drew University ~ November 19, 2009

Analysis of Microtubule Dynamics Using Growth Curve Models:

Microtubules are part of the structural network within a cell's cytoplasm, providing structural support as well as taking part in many of the cellular processes. A large body of data provide evidence that dynamics of microtubules in a cell is responsible for the performance of many critical cellular functions such as cell division. In this article, we study the effect of four different isoforms of a protein tau on microtubule dynamics using growth curve models. The results show that a linear growth curve model is sufficient to explain the data. Moreover, we find that a mutated version of a 3-repeat tau protein has a similar effect as a 4-repeat tau protein on microtubule dynamics. The latter findings conform with the biological understanding of the effect of the protein tau on microtubule dynamics.

Md. Aleemuddin Siddiqi, Ph.D., Symbiance, Inc., Princeton Junction, NJ ~ December 3, 2009