Final Project for Data Science at General Assembly
This purpose of this project is to model the the likelihood of receiving a degree or finding employment in a STEM (science, technology, engineering, and math) field using logistic regression, and derive interesting conclusions from these models.
For a non-technical overview of the project and its results, read my most recent blog post on the project: https://michaelinkles.wordpress.com/2016/04/03/modeling-inequality-in-stem-2/
In order to run any code from this repo, the 2013 American Community Survey data must be downloaded from https://www.kaggle.com/census/2013-american-community-survey to the STEM-degrees folder.
This repo contains the following:
-
stem_low_memory.py: The most recent and most efficient version of the code, but lacks some of the detail of the research process that's in the iPython notebook
-
finalproject-inkles.ipynb: an iPython notebook containing all of my code along descriptions of the code in markdown, documenting my entire process
-
stemdata.py: a python script for all of the data wrangling that needs to be done on the original data for the models to work. it outputs a file called df_formatted.csv which is used by stemmodels.py
-
stemmodels.py: a python script that uses the output of stemdata.py to generate the models and output all of the graphs that are used in my blog post
-
images folder: contains the output of stemmodels.py; all the images that are used in the blog post
-
shapefiles folder: shapefiles to be used by baseplot in generating the map of odds ratios
-
seri.csv: an additional data file containing each state's SERI (Science and Engineering Readiness Index) score, to be used with 2013 ACS data in the data wrangling step.
-
exploratory_interactive.py (new): Code for an interactive visualization of my exploratory analysis that displays a line graph of science degree rate by age. Results can be sorted by state, and grouped by race or sex. This code makes use of Bokeh plotting library, which can be used to create D3.js style graphics in Python. I'm working on getting this running as a standalone web app, but for now the visualization can be viewed by taking the following steps:
- Run stemdata.py
- Install bokeh by typing "conda install bokeh" or "pip install bokeh"
- Type into the command line:
bokeh serve --show exploratory_interactive.py