Collaborators: Raheed Malik
Team: Early Childhood
Goal: Visualize zipcodes with low access to quality child care and how it correlates the the area's race and income
Audience:

A dataset of child care centers for GA, NC, MN, IL, VA, MD, OH, CO, and WA with quality (QRIS) ratings. Each state has a different rating, so we normalize to 3 tiers (top, middle, and bottom) with a 4th category that simply qualifies the center as "licensed." These centers will be overlayed on top of a base layer of zipcode boundaries, shaded by zipcode demographic characteristics.

For the base layer, we’ve decided on four metrics, using a toggle or dropdown:

• Population density using the 2010 Census file in the dropbox + a dataset that has square miles for each ZCTA in 2010 Census.
• Mean Reported Earnings using the IRS data, dividing the amount of salaries and wages in AGI (Excel: column J) by the number of tax returns (Excel: column I) in each ZCTA.
• Percent African American using the 2010 Census ZCTAs
• Percent Hispanic/Latino using the 2010 Census ZCTAs

User should be able to toggle between zipcode characteristics and hover over a center to see more details.

• Requirements
• Quick start
• Project Architecture
• Performance Concerns
• Work Remaining

• node ^4.2.0
• npm ^3.0.0

• Clone this repository git clone https://github.com/amprog/childcaredeserts.git
• Install dependencies npm install
• Running npm start will start a dev server on port 8080

Child Care Deserts development

Within the download you'll find the following directories and files, logically grouping common modules.

childcaredeserts/
├── bin/
│   ├── munge
│   ├── topo
├── public/ // resources like images, csvs, and geoJson files
│   └── data/
│       ├── zips/ // all zip code topojson and geojson files
│       │   ├──  zips_*.geo.json
│       │   └──  zips_*.topo.json
│       └── childCareCenters.tsv // data for each child care dot
├── js/
│   ├── api/ 			// AJAX related code
│   ├── ui/ 			// React-Redux related code
│   ├── index.js		// Main entry point. Creates Redux store and React root node
├── sass/    // SASS modules for typography, colors, layout, etc
└── views/   // Mustache-related views used by build system
   

This project primarily uses React, Redux, and Mapbox GL to load, filter, and render geospatial data.

The ui/ directory divides the app into to primary parts: sidebar/ and map/. The components and containers located in sidebar/ do most of the heavy lifting for filtering, selection, and loading data.

map/ uses components to divide the MapboxGL map into a map, a camera, and layers. Panning and zooming to features is done by MapboxGl.component.camera.js, while the various data layers (e.g. Child Care locations and state ZIP codes) are loaded and removed via MapboxGl.component.layer.js.

All state changes are made through Redux and provided to each React component via react-redux connect.

Displaying thousands of data points comes at a performance price. This section outlines performance bottlenecks encountered, and what we did to ameliorate them.

The project's data for Child Care Centers childCareCenters.tsv is 2.8MB alone. Other files, such as the *.geo.json files located in zips/ can be from 1MB to 5MB in size. Use of gzip when sending files over the wire is highly recommended.

Mapbox uses a thing called a data source to create visual layers. If the size of the source is too large, performance takes a significant hit. In order to reduce the impact of rendering 25k child care centers, we downsampled the size of the data set when viewed at the national view. When a user selects a state, we replaced the child care center source with the all the state's child care centers.

For ZIP codes, topojson is a natural fit because of its file size. However, decompressing topojson files on the client and loading them into Mapbox GL proved to have a significant performance hit. We decided to use urls instead, which allow Mapbox GL to load everything via web workers. However, all sources must be geojson instead, which comes at a significantly larger size. mapshaper was employed to find a compromise between file size and visual fidelity.

Due to changing assumptions driven by performance, several parts of work remain.

The data for each zip code is stored as geojson. However, the actual points of data, i.e. mean earnings, population density, percent African-American, and percentage Latino are either not present, or described as non-numeric strings or null. This causes Mapbox GL's data-driven layers to fail. All zip code geojson data must be updated and must pass quality assurance from relevant stakeholders.

Currently, only Colorado is working for ZIP codes.

1. Ensure all ZIP code properties are numeric and not strings
2. Ensure all ZIP code properties are non-null
3. Ensure all ZIP code properties have consistent units accross all states (e.g. thousands vs tens)

The easiest way to do this is to update either a csv or an excel document, and then use topojson's console api to marry the zip code in the spreadsheet to the zipcode in the geojson.

Typography, layout, have not yet been built into this project. This project is therefore only a technical proof of concept and is not ready for audiences.