Department Colloquia

Annie Phung

A Random? Walk Down Wall Street, Part I

Time series analysis plays a unique role in the field of Data Science. In the financial industry, it has become a powerful tool for predicting future stock price movements. We explore how investors can take advantage of the lack of independence in the rate of return process over time and examine the application of different time series models.

• Done in partial fulfillment of the Financial Mathematics major.

Presentation Details:

• November 15, 2022 at 4:00pm
• This will be held online using Zoom.
• Advisor: Kevin Hastings
  Please contact the advisor for more information about this presentation.

Ashus Owaisi

A Random? Walk Down Wall Street, Part II

In this study we use AR, MA, ARMA and ARIMA models to look at a random sample of companies from the Fortune 500 list for 2022. Using the aforementioned models and techniques such as the autocorrelation, partial and extended autocorrelation, decomposition, forecasting and residual analysis, we will uncover the truth about their rates of return. These rates may be a lot more predictable than we previously thought.

• Done in partial fulfillment of the Financial Mathematics major.

Presentation Details:

• November 15, 2022 at 4:00pm
• This will be held online using Zoom.
• Advisor: Kevin Hastings
  Please contact the advisor for more information about this presentation.

Anastasiia Ganshina

Grading individuals based on American Political Landscape: A Wikipedia exercise

It is well known that there are two main opposing parties in the United States of America today: Republicans and Democrats. Both parties have their solutions to major topics in all sectors of life of Americans and the world, and, in many cases, those opinions are contradictory. Most people in those parties, however, do not always fully agree with their party rules and opinions. For example, they might have most of their opinions aligned with republicans, but still have some ideas that are democrat leaning. For instance, they may be pro-choice and pro-gun but believe in universal healthcare. In this study, our main goal was to create a system that would grade politicians based on their political opinions on a scale from 0 to 10. We studied a few different topics, such as abortion, guns, healthcare, etc. to find out what the democrats and republicans think about them and grade a politician based on opinions. For our grading system, we developed over 50 machine-learning language-based models that can determine the sentiment based on a portion of the Wikipedia article on a specific topic.

• Special presentation for the Statistics Program

Presentation Details:

• November 10, 2022 at 4:00pm
• Room: SMC A-203
• Advisor: Ole J. Forsberg
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Ashus Owaisi

The Japanese Warring States: A Statistical Study

The Warring States period, also known as “Sengoku Jidai” is the period known within the archipelago between 1467-1615. An era known for its brutality and heroic displays thanks to constant social upheaval, civil wars, and power struggles. This study attempts to show a historical analysis of all the major events that transpired during this period. The nation was unified three separate times, under three vastly different ruling entities. During which a multitude of clans took part in shaping the history of the modern nation, as we know it today.
 Battles, Army Counts, Battle Locations, Sieges, Naval Tactics, Battle Instigations, Forced Defenses, Japanese Prefectures/Regions, and Political Conspiracies are all taken into account in this study. Using regression and prediction analysis techniques, we produce this statistical and historical recap and take a look back at one of the most infamous periods in time.

• Done in partial fulfillment of the Statistics minor.

Presentation Details:

• May 26, 2022 at 4:00pm
• Room: SMC A-203
• Advisor: Ole J. Forsberg
  Please contact the advisor for more information about this presentation.

Ole J. Forsberg

Beyond Frequentist Statistics: A brief introduction to Bayesian statistics

The vast majority of undergraduate statistics follow the frequentist framework as conceived by Ronald Fisher (18-19) and others. This framework leads to mental gymnastics when handling such foundational concepts as confidence intervals and p-values. This talk briefly introduces Bayesian statistics and how it allows for a much cleaner interpretation of the data.

• Special presentation for the Statistics Program

Presentation Details:

• May 24, 2022 at 4:00pm
• Room: SMC A-203

Beck Baird

What Is "Not Statistically Significant:" Bayesian Insight on Data Analysis

Data was collected longitudinally through surveys completed daily for two weeks by participants in an experimental or a control group. The aim was to find a significant correlation between experimental condition, self-determination, and self-concept clarity. Frequentist analysis suggested that manipulation was not statistically significant, but further Bayesian analysis gives a stronger insight into the meaning of our findings.

• Special presentation for the Statistics Program

Presentation Details:

• May 24, 2022 at 4:15pm
• Room: SMC A-203
• Advisor: Ole J. Forsberg
  Please contact the advisor for more information about this presentation.

Samantha Lorenz

Testing Differential Invalidation with Frequentist and Bayesian Analysis

Democratic backsliding has recently become a trend. In Hungary's case, one party seems to have triggered this event: the Fidesz party. In this study, I examine the changed electoral system and study the effects of differential invalidation for the Fidesz party in order to make a conclusion on whether or not there is evidence of differential invalidation in the case of Democratic Backsliding. I have taken an initial project a step further and evaluated the data both with Frequentist and with Bayesian methods. My goal is to go over the differences between the two types of analysis, and which one is better.

• Special presentation for the Statistics Program

Presentation Details:

• May 24, 2022 at 4:30pm
• Room: SMC A-203
• Advisor: Ole J. Forsberg
  Please contact the advisor for more information about this presentation.

Anastasiia Ganshina

Surgical Mask Detection

In 2020, the world was shaken by COVID-19, the pandemic that continues to affect our lives. Due to mutations and constant challenges humanity is facing, this illness will stay in our day-to-day lives for some time. In order to limit the spread of the disease and keep each other safe, CDC primary recommendation is to wear masks in public places or when closely interacting with others. However, there is no good way to monitor whether these rules are followed and people are protected. Thus, the creation of the system of autonomous mask detection might be helpful to make sure people are keeping each other safe and healthy.
 The autonomous system described above can be simply implemented by using Machine Learning and, more specifically, the Computer Vision component of the field. Computer Vision is a part of the Machine Learning field that deals with images and videos, and bases the predictions from feature extraction of patterns detected in images or frames. Thus, to build the mask detection model, the images of peoples' faces with and without masks are needed.
 This research project creates such a system using freely available materials, including Python and Keras. The accuracy of this model is explored and explained.

• Done in partial fulfillment of the Data Science major.

Presentation Details:

• March 1, 2022 at 4:00pm
• Room: SMC A-203
• Advisor: Ole J. Forsberg
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Dieu Kim (Zoey) Nguyen

Analyzing the Olympic Record

The modern Olympic Games have been held, with three exceptions, every four years since 1896. Much has changed from Coubertin's inaugural 1896 Olympics in Athens, which had just 241 athletes competing in 43 events. The 2020 Tokyo Olympics had 11,656 athletes competing in 339 events. Each event offered a gold, silver, and bronze medal.
 This research uses several data science techniques to determine the effects of height, weight, and age on the likelihood of an athlete medaling in an event. Several events, from both Summer and Winter Games, are examined. Results are explored using data visualization.

• Done in partial fulfillment of the Data Science major.

Presentation Details:

• March 3, 2022 at 4:00pm
• Room: SMC A-203
• Advisor: Ole J. Forsberg
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Hao Li

Using neural networks to recognize handwritten digits and to approximate simple functions.

The problem we are going to solve is how to use neural networks to recognize handwritten digits. This presentation will focus on constructing the network and understanding the mathematical principles behind the network. The network will be implemented by Python code and through a little modification, such neural network can be used to approximate simple multivariable functions

• Done in partial fulfillment of the Mathematics major.

Presentation Details:

• May 28, 2021 at 4:00pm
• Room: SMC A-203 and Zoom
• Advisor: Andrew Leahy
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Libby Winchester

Biases in the Great British Baking Show

In the weekly episodes of the Great British Baking Show (GBBS) there is a “technical challenge” round where the contestants’ final products are judged blind, and then ranked from worst to best. This research looks into how several cognitive biases play a role in the judging process, affecting the outcome of the challenge.
 If and when these biases are in effect, then the placement of the bake in the taste test affects its ranking. This information is important to recognize and notice how much cognitive biases play into people’s responses to various propositions, and how that might change the work of researchers and statisticians in the future.

• Done in partial fulfillment of the Statistics minor.

Presentation Details:

• May 27, 2021 at 4:00pm
• Room: SMC A-203 and Zoom
• Advisor: Ole J. Forsberg
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Elliot Bainbridge

Research into the use of advanced statistical models to forecast asset options prices

The focus of this research is to investigate the use of advanced statistical models, that have been developed during the 21 st century along with fast modern computer programming language, to create a model that accurately maps and forecasts market options prices. The forecasting of this model will provide vital information for the financer as to whether an option is under or overpriced, something that the BSM model does not allow. Throughout this research we will be developing our knowledge of options and their associated terminology, Python computer programming, advanced statistical modelling, and forecasting.

Presentation Details:

• May 26, 2021 at 4:00pm
• Room: SMC A-203 and Zoom
• Advisor: Ole J. Forsberg
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Jason White

Interdependency of Random Rates of Return

In this research, Jason investigates the independence assumption of Geometric Brownian Motion. He will share with us the various tests he conducted on a variety stocks throughout different years including the Chi-Square Independence Test, Student t distribution, and Runs test. There will also be a section investigating the stationarity of these rates of return based upon Autocorrelation Functions and Partial Autocorrelation Functions.

• Done in partial fulfillment of the Financial Mathematics major.

Presentation Details:

• March 11, 2021 at 4:00pm
• This will be held online using Zoom.
• Advisor: Kevin Hastings
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Zoey Nguyen

Is There a Better Model for Predicting Stock's Rate of Return?

The research will be about testing to see whether stock's rate of return actually follows a normal distribution that a Geometric Brownian Motion assumes. By looking at many different stocks, we can see where normal distribution might fail and thus, propose other alternate distributions that might perform better. Then, I will be using three goodness of fit tests: Jacque-Bera, Kolmogorov-Smirnov, and Pearson's Chi Square to verify my findings.

• Done in partial fulfillment of the Financial Mathematics major.

Presentation Details:

• March 11, 2021 at 4:00pm
• This will be held online using Zoom.
• Advisor: Kevin Hastings
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Elliot Bainbridge

Forecasting Tesla's Opening Stock Price using Advanced Statistical Model Building Techniques

This research uses advanced statistical model building techniques to forecast the opening stock price of Tesla, Inc., on 2 November 2020. Modelling techniques used include: multivariate linear models, exponential smoothing methods, ARIMA modelling, and Vector Auto-Regressive (VAR) models. Each method provides different ways to forecast the price of Tesla on that day.
 Each model also gives some insight into stock prices. To determine which model was superior, the predicted and actual price of Tesla were compared. Surprising results arose. Come find out what they told us!

• Done in partial fulfillment of the Statistics minor.

Presentation Details:

• February 25, 2021 at 4:00pm
• Room: SMC A-204 and Zoom
• Advisor: Ole J. Forsberg
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Joshua Schumacher

Introduction to Commodity Swaps, Interest Rate Swaps, and Currency Swaps

This presentation will focus on commodity swaps, interest rate swaps, currency swaps, and arbitrage pricing, reducing the confusing financial jargon, in a way more suitable for the average person to understand. We will also expand on commonly used financial terms, such as forward contracts and arbitrage.

• Done in partial fulfillment of the Financial Mathematics major.

Presentation Details:

• February 17, 2021 at 4:00pm
• This will be held online using Zoom.
• Advisor: Kevin Hastings
  Please contact the advisor for more information about this presentation.
• Please download the flyer for more information.

Ole J. Forsberg

 Improving Electoral Estimates: Leveraging Available Information 

The United States elects its president with the Electoral College, a body of individuals elected by each state by the people of that state. This leads to some interesting results, especially when the national popular vote does not align with that of the Electoral College. Most tracking polls available to the people focus solely on the national vote, letting us know who is ahead and by how much… but only at the national level. It is the state vote that counts, however. 
 In this talk, I briefly explain my polling model in which I use a modified hierarchical time series model to estimate candidate support at the state level, thus allowing us to estimate the winner of the Electoral College over time. The results of the model align nicely with past elections. The model also allows us to discuss changes in support level over time and the likelihood of a Trump victory in 2020. 

• Special presentation for the Statistics Program

Presentation Details:

• October 30, 2020 at 4:00pm
• This will be held online using Zoom.
• Please download the flyer for more information.