This project is focused on identifying key factors behind food insecurity in the United States and determining their relative importance via multiple regression analysis. 

Food insecurity is "a household-level economic and social condition of limited or uncertain access to adequate food" (USDA, 2015), and is an increasingly common phenomenon across the United States. Many factors, such as the increase in food prices and increase in population, put increasingly high burden on the ability of American families in lower socioeconomic classes to find food. Data for these various factors is readily available in the Food Environment Atlas by the USDA, the US Census, and the Council for Community and Economic Research, and includes data on each state. For this project, I specifically looked at the percentage increase in grocery stores from 2007 to 2012, the percentage increase in population from 2007 to 2012, and the percentage increase in the grocery index (a component of the cost-of-living index) from 2007 to 2012. The goal of this project is to determine which of these factors impacts food insecurity most; that is, a 1% change in which factor will lead to the maximum decrease in food insecurity?

The data processing pipeline for this project involved 3 parts: 1) data sanitization and joining, 2) multiple regression analysis of data set, 3) display of results. In the data sanitization and joining process, multiple data sets including information on food, population, and cost of living index were identitied, sanitized (e.g. filling in of missing values), and joined based on state using Excel. In some cases where direct percentage change data was unavailable, the percentage change was calculated by looking at 2007 and 2012 values. Once this process was completed, the data was converted to tsv form and parsed into a Node.js based parser. The data was then run through a standard multiple regression analysis program (written by me in Node.js), and the final equation representing the regression function was returned. Since the data was normalized to use only percentage increases in each factor, the magnitude of the coefficients in this case represent the relative importance of each factor in determining food insecurity. The results were returned in two forms: 1) the final regression function, 2) a list of the most imperative issue in each state, as shown in the map.

The final regression function was y = -0.048 * x1 + 0.183 * x2 + 0.062 * x3 + 0.431, where: 
	y = % increase in food insecurity
	x1 = % increase in grocery stores
	x2 = % increase in population
	x3 = % increase in grocery index.
As shown, the increase in population appears to have the most significant effect on food insecurity.

As a result of using only a few limited variables to explain a highly complex issue like food insecurity, the accuracy of the existing model is limited. However, the accuracy of this model can be improved by adding more variables, like average distance to grocery stores, to the list. 

The code for the multiple regression and matrix JS libraries are sourced from the author at https://github.com/ShreyGupta19/multipleregression and https://github.com/ShreyGupta19/matrixjs.