2 Case Study: Health Effects of the Deepwater Horizon Oil Spill
On the evening of April 20, 2010, the Deepwater Horizon, an oil drilling platform positioned off the coast of Louisiana, was engulfed in flames as the result of an explosion. The drilling rig, leased and operated by BP, had been tasked with drilling an oil well in water nearly 5000 feet deep. Eleven personnel were killed in the explosion. The following screenshot is from the initial coverage by the New York Times1:
The incident is considered the worst oil spill in US history, creating an environmental disaster along the Gulf Coast. In addition to studying the effects on the local environment, researchers have undertaken studies to examine the short and long-term health effects caused by the incident. As an example, we might ask whether volunteers who were directly exposed to oil, such as when cleaning wildlife, are at higher risk of respiratory irritation compared to those volunteers who were helping with administrative tasks (and therefore were not directly exposed to oil). An article appearing in The New England Journal of Medicine (Goldstein, Osofsky, and Lichtveld 2011) reported the results from a health symptom survey performed in the Spring and Summer of 2010 by the National Institute for Occupational Safety and Health. Of 54 volunteers assigned to wildlife cleaning and rehabilitation, 15 reported experiencing “nose irritation, sinus problems, or sore throat.” Of 103 volunteers who had no exposure to oil, dispersants, cleaners, or other chemicals, 16 reported experiencing “nose irritation, sinus problems, or sore throat.”
While a larger fraction of volunteers cleaning wildlife in the study reported respiratory symptoms compared to those who were not directly exposed to irritants, would we expect similar results if we were able to interview all volunteers? What about during a future oil spill? Is there evidence that more than 1 in 5 volunteers who clean wildlife will develop respiratory symptoms? What is a reasonable value for the increased risk of respiratory symptoms for those volunteers with direct exposure compared to those without?
In the first part of this text, we use this motivating example as the context for discussing how research questions should be framed, methods for data collection, summarizing and presenting data clearly, quantifying the variability in an estimate, and quantifying the degree to which the data disagrees with a proposed model. We capture these ideas in what we call the Five Fundamental Ideas of Inference. We will also see that any statistical analysis moves between the components of what we call the Distributional Quartet. These two frameworks allow us to describe the language and logic of inference, serving as a foundation for the statistical thinking and reasoning needed to address more complex questions encountered later in the text.