index_creation.Rmd

---
title: "Creating Marginalization and Socioeconomic Hardship Index"
author: "Census Team"
date: "3/9/2022"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

We decided to create an index to allow people interested in the topic to get a 
quick overview of some important demographic factors when looking at Divvy data.
Working with our mentor and using his paper "Investigating socio-spatial 
differences between solo ridehailing and pooled rides in diverse communities" as
a jumping off point.  We variable correlations and used Exploratory Factor Analysis
with a factor loading cutoff of .3, a single factor, and no rotations to create an index. 

Factor analysis is a way to reduce data, so instead of having to look at 10 
different heatmaps, each colored by a single demographic variable, we use factor 
analysis to create an index that represents all 10 variables with a single number.  
We use demographic variables that represent various aspects of marginalization 
and/or socioeconomic hardship (like the percent of population that is disabled 
or the unemployment rate) and used a factor loading cutoff of .3.  
Factor loadings measure how strongly related each of our demographic variables 
is to our index variable and range from -1 to 1.  The sign on the factor tells 
us the direction of the factor variable’s relationship with our overall index 
variable, positive is they tend to rise and fall together (like our index and 
percent of families living below the poverty level) and negative is they rise 
and fall in opposite directions (like our index and median income).

The selected variables were:

-Percent of the population that is not non-hispanic/latino white
-Percent of the population that is disabled
-Percent of the population that do not have access to a vehicle
-Percentage of households that are renting
-Percentage of households with a single parent
-Percentage of households with no internet access
-Unemployment Percentage
-Percentage of the population with a bachelors degree
-Percentage of families living below the poverty line
-Median income


Some variables we tried that were not selected due to low factor loadings were:

-Percentage of the population that do not speak English "very well"
-Percentage of commuters that commute by car, truck, or van

We used this factor analysis to create an index and scaled the index between 0 and 1
to be used in our analysis and visualizations.