Ryan Whetstine 501 Statistical Literature and Problems Project Presentation

Tuesday, December 8, 2020 - 1:00 PM - 2:00 PM
Contact

Kathie Leck
Fariborz Maseeh Department of Mathematics + Statistics

Ryan Whetstine will give a talk on the following statistical literature and problems project
in partial fulfillment of requirements for the

Master of Science in Statistics 
    
Topic:

“A Simple Parametric Model Selection Test”


This project is based on a paper titled " A Simple Parametric Model Selection Test" by authors Susanne Schennach and Daniel Wilhelm. 

We examine a simple model selection test for choosing between two parametric likelihoods in the following scenarios: both models, one, or neither are allowed to be correctly specified or misspecified. They could be nested, non-nested, strictly nested, or overlapping. No pre-testing is required, since in each case the same test statistic along with a standard normal critical value can be used. This procedure controls for asymptotic size uniformly over a large class of data-generating processes. We demonstrate the test's finite sample properties in a Monte Carlo experiment and its practical relevance in an application comparing Keynesian vs new classical macroeconomic models

Under the direction of 
Dr. Nadee Jayasena
 

Upcoming events

Urmas Yanis 501 mathematical literature and problems project…

Urmas Yanis will give a talk on the following mathematical literature and problems project in partial fulfillment of requirements for the Master of…
Add to my Calendar 2020-11-25 14:00:00 2020-11-25 15:00:00 Urmas Yanis 501 mathematical literature and problems project presentation Urmas Yanis will give a talk on the following mathematical literature and problems project in partial fulfillment of requirements for the Master of Science in Mathematics Topic: “Overview and Demonstration of Supervised Deep Learning” Abstract: In this 501 project we shall give an overview of supervised deep learning by presenting three algorithms used in deep learning: backpropagation, used for computing a neural network's gradients, and the batch and stochastic gradient descents, used for updating a neural network's weights and biases. The aforementioned backpropagation algorithm will be the core of this presentation. We'll also demonstrate how a feedforward neural network coded in Python 3, not relying on Keras or TensorFlow, can achieve results above 95 percent of the MNIST and Fashion-MNIST databases. Under the direction of Dr. Mau Nam Nguyen   https://pdx.zoom.us/j/82904292252 Kathie Leck Fariborz Maseeh Department of Mathematics + Statistics Kathie Leck Fariborz Maseeh Department of Mathematics + Statistics America/Los_Angeles public

Mitchell Fennimore 501 statistical literature and problems…

Mitchell Fennimore will give a talk on the following statistical literature and problems project in partial fulfillment of requirements for the…
Add to my Calendar 2020-11-30 10:00:00 2020-11-30 11:00:00 Mitchell Fennimore 501 statistical literature and problems project presentation Mitchell Fennimore will give a talk on the following statistical literature and problems project in partial fulfillment of requirements for the Master of Science in Statistics  Topic: “The resolution of a fractional factorial experiment in terms of error correcting codes” Abstract: Based on: “An application of information theory and error-correcting codes to fractional factorial experiments” Journal of statistical planning and inference, 92 (2001) Irad Ben-Gal, Lev B. Levitin An isomorphism between the fractional factorial experiment and error correcting codes helps unravel the concept of a design’s resolution and the logic of alias structures. Once a resolution is decided on, inequalities from the study of error correcting codes provide optimality criteria for design selection. This criterion leads to a “paradoxical,” outcome: The resolution of a fractional factorial design can be maintained with a smaller experiment by increasing the number of levels.  Under the direction of  Dr. Robert Fountain https://pdx.zoom.us/j/89263044111 Kathie Leck Fariborz Maseeh Department of Mathematics + Statistics Kathie Leck Fariborz Maseeh Department of Mathematics + Statistics America/Los_Angeles public

Aaron Maroni 501 mathematical literature and problems project…

Aaron Maroni will give a talk on the following mathematical literature and problems project in partial fulfillment of requirements for the Master of…
Add to my Calendar 2020-12-04 13:00:00 2020-12-04 14:00:00 Aaron Maroni 501 mathematical literature and problems project presentation Aaron Maroni will give a talk on the following mathematical literature and problems project in partial fulfillment of requirements for the Master of Science in Mathematics Topic: “A Discussion on Reviving the Method of Particular Solutions in Solving Laplace Equation's Eigenvalues” Abstract: In this project, I will present a research paper published in SIAM Review, "Reviving the Method of Particular Solutions (MPS)" written by Betcke and Trefethen. This paper first addresses the deficiency of the original method from "Approximations and Bounds for Eigenvalues of Elliptic Operators" written by Fox, Henrici and Moler. It then proposes an improved MPS which allows computation of eigenvalues of the Laplace's equation with Dirichlet boundary conditions for much better accuracy over the original MPS. Furthermore, it also enables treatment of domains with multiple singular corners. Therefore, it has dramatically expanded the capability and robustness of the MPS method. I will present the work given by Betcke and Trefethen, and then develop new Python code to solve eigenvalues of the Laplace's equation on the domains with multiple singular corners. The new Python algorithms achieves even better performance than those reported in the original paper which unfortunately doe not share their code. The new Python code is thus attached in the Appendix for open access by other researchers. Under the direction of Dr. Bin Jiang https://pdx.zoom.us/j/81597994369 Kathie Leck Fariborz Maseeh Department of Mathematics + Statistics Kathie Leck Fariborz Maseeh Department of Mathematics + Statistics America/Los_Angeles public