Introduction

The food system in the United States has grown more and more complex since the 1930s. Many factors, including large crop subsidies, changing methods and prices of transportation, the cost-effectiveness of large-scale agribusiness, billions spent on advertising, and effective industry lobbying, have contributed to these complexities. 
Although many nutrition choices may seem individual, there are many influences on our access, quality, availability, and affordability of various types of food. In America today, it is far cheaper to purchase a hamburger at a fast-food restaurant than to buy produce at a gracery store. Access to a grocery store in the first place is often limited, in what have been termed "food deserts". These are areas, often in urban centers or in rural locations, where someone at median income is more than a certain parameter (either distance or time) from an affordable full-service grocery store. Generally, the guideline is more than one mile away from one's home.

Food deserts most negatively impact groups such as the elderly, low-income, and anyone else who may not have access to a personal vehicle. Living in food deserts often means choosing to obtain food from local corner stores or fast service restaurants – which can be more expensive and often less healthy. Lacking access to healthy food has severe consequences. Around the country, more than one in seven people is considered food insecure. Whereas hunger means a direct lack of enough calories, food insecurity means lacking the resources to provide a reasonably healthy, adequate diet for a normal lifestyle. People who are food insecure may sometimes go hungry, but more often find themselves choosing food they know is unhealthy instead. This is the link between poverty and obesity in the U.S. When your food budget is limited, often you will make choices to eat food that is not nutritious. Obesity and malnutrition in America have their own consequences. Over the past two decades we have seen rates of childhood and teenage obesity skyrocket, as well as rates for adults. Risk of chronic diseases such as diabetes and heart disease are much higher for obese individuals. And the collective costs these bring to healthcare are a major contributor to that sector’s rise in prices. Food Swamps are areas that have high concentrations of fast food restaurants. They are often, though not always, seen in conjunction with food deserts. A combination of lack of access to healthy food and readily available unhealthy food makes it extremely difficult for many to maintain a healthy lifestyle. Not surprisingly, these areas hit low-income areas the hardest, and become concurrent issues with poverty in the realm of public health

Study

I wanted to examine the food environment in the United States by looking at the relationship between two variables – obesity and fast food. I created a bivariate map to show concentrations of obesity, of fast food, and of the range in between. In this study, there were two independent variables – fast food per capita and obesity rates. The unit of analysis was US county, determined by FIPS code.

Method

-Data collection Both of my variables were found through the United States Department of Agriculture, in a data set titled “Food Environment Atlas” released on April 3rd, 2015. The dataset can be found here: https://catalog.data.gov/dataset/food-environment-atlas-f4a22 Obesity rates for adults in 2010 is in the “Health” sheet, and Fast Food Restaurants per 1,000 people in 2011 is in the “Restaurants” sheet.

-Data Cleaning I created a dataframe using the two variables and FIPS codes for counties. I removed territories, Alaska, and Hawaii.

-Bivariate Analysis method In order to see the relationship betwen the two variables using spacial data, I needed to arrange my data into nine “buckets” – each with a corresponding color. Each variable would have a different color scheme, and the colors in the middle would be the relatioship between the two. I cut each variable into three segments. -The range of obesity rates was 13.1 to 47.9, with intervals of 11.6. -The range of fast food/1000 people was 0 to 5.79, with intervals of 1.93. In Excel, I created two columns, each with an integer 1, 2, or 3 corresponding to the interval segment for each variable. The combination of these integers corresponds to a high-mid-low level for the two variables together.

Project.Data <- read.csv("Project.Data.csv")
head(Project.Data)
##   X FIPS  FFRPTH11 PCT_OBESE_ADULTS10 Col2Bucket Col3Bucket   AD    B
## 1 1 1001 0.6151953               30.5          1          2 TRUE TRUE
## 2 2 1003 0.6480395               26.6          1          2 TRUE TRUE
## 3 3 1005 0.7006158               37.3          1          3 TRUE TRUE
## 4 4 1007 0.2635509               34.3          1          2 TRUE TRUE
## 5 5 1009 0.3467587               30.4          1          2 TRUE TRUE
## 6 6 1011 0.3794346               42.1          1          3 TRUE TRUE
##     color
## 1 #E88D55
## 2 #E88D55
## 3 #FFB35E
## 4 #E88D55
## 5 #E88D55
## 6 #FFB35E

For example, a “1,3” would mean a county has a low rate of fast food, but a high rate of obesity. A “3,1” would mean the inverse.

I then imported my data into R and used the following code to assign a color to each possible combination. I repeated this nine times until each combination had an assigned color.

#Project[(Project$Col2Bucket == 1) & (Project$Col3Bucket==1),"color"] <- "#FF9966"

#I needed to change some of my colors later on, so I used code such as the following:
#Project3$color <- as.character(Project3$color)
#Project3$color[Project3$color == "#fff685"] <- "adar"
#Project3$color[Project3$color == "#B59585"] <-"#FF9966"
#Project3$color[Project3$color == "adar"] <- "#fff685"
#Project3$color <- as.factor(Project3$color)

Once I had this completed dataset, with each county being assigned a color, and each color having a meaning, I could start mapping.

Map

I used the maps and mapsproj packages to write code that would allow me to connect counties using FIPS identifiers. 
-First, the package needs colors assigned. 
-Then, there is a step where the colors that were just assigned are matched with the color vector and built-in FIPS codes. 
-Next is mapping out at a county level, inserting the colorsmatched object, and using arguments for the package (such as resolution, projection, etc). 
-I added white state borders by overlaying the states map, but keeping the fill as FALSE, and I increased the size of county borders. 
library(maps)
library(mapproj)
Project3 <- Project.Data
data(county.fips)

Findings -The main color/theme (light orange) is the “1,1” or low fast food and low obesity rates. However, there are two regions that stand out. On the western part of the country there is a bright yellow, which indicates low fast food rates and medium obesity rates (“1,2”). These areas have less than 2 fast food restaurants per one thousand people, and obesity rates between 24.7 and 36.3. The other notable color is bright purple, which indicates low fast food restaurants and high obesity rates (between 36.3 and 47.9).

-There were only nine counties that had more than "low" rates of fast food restaurants per capita. In particular, one county in Colorado had 5 fast food restaurants for a total population of 780, which may have skewed the data.

Limitations -I used fast food data from USDA under one category. However, there are many food service places that convey unhealthy food which may have not been counted in that particular category. I think that if I had better access and resources, I would have liked to do a more in-depth counting of unhealthy food services. -Many of the programs to map in R do not use FIPS codes, and the maps package does not allow for many features which would have been helpful, such as easily placing city names or labels and annotations.

Discussion While this map does not show a full picture of the food environment in the United States, it shows that some areas are more prone to obesity than others. In particular, parts of the south and west are likely to have higher rates of obesity. This map does not show a particularly strong relationship between fast food and obesity, though I believe that with better data and abilities, I would have been better able to show that relationship.