Philip Bradfield - projects from DC-DSI-3 (2016-2017)
Ground Level Ozone - Modeling ground level ozone with data from the EPA. Ground level ozone is an atmospheric pollutant that is formed by chemical reactions between oxides of nitrogen (NOx) and volatile organic compounds (VOC) in the presence of sunlight. Several different modeling approaches were used to predict if the level of ozone was above a threshold. They included SVM, Logistic Regression, and various decision tree methods were used.
West Nile Virus in Chicago - An analysis of mosquito traps in Chicago (team project). Worked with two exceptional women from the DC-DSI-3 cohort, Ritika and Parisa. Initial research I did found that the length of day was an important variable so we included it in our model. Our team won 1st place for best presentation.
Iowa Liquor Sales - Make a projection of 2016 sales based on model built on data from 2015. The data on liquor sales came from the state of Iowa. We were given a listing of every wholesale purchase of alcohol (not beer or wine) for Iowa in 2015 and the first quarter of 2016. I used a linear regression approach and built a model for the 1st quarter of 2015 with the target being the total sales for 2015. I then used that model and the 1st quarter of 2016 to project sales for 2016.
United States SAT Scores - Analyze and explain observed average state SAT scores. The data was composed of the SAT (math and verbal) scores for each state (and DC). I analyzed the data and presented a strong case that states where the ACT is prevalent only the top students take the SAT and that affects the average SAT scores for those states.
Global Terror Database - Bayesian comparison of France and UK bomb attacks. The Global Terror Database (GTD) is a collection of data about terror attacks that have occurred anywhere in the world from 1970 till 2015. A Bayesian analysis was done comparing the number of terror related bombings in the UK and France. Then, an brief analysis on the number terror related deaths of USA citizens was done. Finally, missing data from 1993 is imputed.
Web-Scraping the Indeed Job Site - Gather/analyze job salary info from the Indeed web site. The Indeed.com job posting web site is one of the major aggregators of science related job openings. Company name, location, job titles, salary information and other relevant information was ‘scraped’ from the Indeed web pages as well as related pages posted by the companies that were looking to hire someone for a data science/analyst related job. Information from more than 1000 job postings was gathered using web scraping techniques. That information was analyzed and used to build two different models that could be used to indicate if the pay for a job would be more than the median income for a data related science job in the DC area.
Bayesian Dice Handout - A handout made up to make it easy to determine the odds of rolling a specific number when dies with different numbers of sides are used. Also, a layout of how to calculate the odds of specific die being used when a specific number is rolled.