UCB Grade Distributions

A tool for exploration of grade distribution data obtained from Cal Answers.

Requirements

Python 2.7.x, Postgres, and the packages in requirements.txt.

Setup

Local Development

• Create a virtualenv and install the Python dependencies with pip install -r requirements.txt.
• In Postgres, create a database for this. For example, if you wanted to call it ucbgradedists, run CREATE DATABASE ucbgradedists.
• Set these environment variables:
  • DJANGO_SECRET_KEY: a secret key (usually a randomly generated string)
  • DJANGO_SETTINGS_MODULE: use ucbgradedists.settings.local if you're in a development environment, and ucbgradedists.settings.production in production.
  • DATABASE_URL: a URL to your database. See the dj_database_url README for the format.
  • ADMIN_NAME, ADMIN_EMAIL: a name and email for the admin of the site.
  • EMAIL_HOST_USER, EMAIL_HOST_PASSWORD: Gmail user and password for the email to send from.
  • SERVER_EMAIL: the email that the server should send emails from (to admins/managers).
• Run python manage.py migrate to run the migrations and create tables for the models.
• Run python manage.py import_csv [season] [year] [inname] on a CSV file to import data. See python manage.py importcsv -h for more information.
• Run python manage.py import_dir [dir] to import data from all CSV files in a directory.
• Run python manage.py import_disciplines to import a mapping of subjects to academic disciplines.
• Run python manage.py compute_stats to compute subject statistics.

Deploying on Heroku

• Set up your app and dyno on Heroku:

    heroku create <your_app_name>
    git push heroku master
    heroku ps:scale web=1
  

• Set the environment variables listed above, either through Heroku's web interface, or with the CLI tool.

• Run heroku run python ucbgradedists/manage.py migrate to create the tables.

• (TODO) Eventually, we'll have an admin/superuser page to import CSV data.

Things You Might Want to Do After a (Big) Commit

• Install new requirements: pip install -r requirements.txt
• Run migrations: python manage.py migrate

Functionality

Current

• Importing grade distributions from Cal Answers CSV output.
• A RESTful API (via Tastypie) to access this data at /api/v1/. Check out the Tastypie docs to figure out how to work with this!
• A basic UI for accessing our data at /search/divisionsets/.
  • We can browse data in terms of DivisionSets, which are sets of divisions (e.g. upper, lower, graduate) on which we compute subject-wide statistics.
  • This includes basic grade statistics -- average, standard deviation, median -- for each subject, course, and section. Thanks to DataTables, it's (mostly) sortable and searchable.

Future

• Things that should be sortable, but aren't: terms, course/section numbers.
• (QUESTION) Which classes/subjects have the highest/lowest averages/medians/stdevs?
• (QUESTION) What do people think are the hardest/easiest (maybe say "highest/lowest grades") classes on campus? Within their major? How well does this match up with reality?
• Put a histogram (or similar chart) for each section (course? subject?).
  • The first two (section, course) are already covered by Berkeleytime and ScheduleBuilder for a lot of classes/terms, and it works pretty well.
  • Choice: grade points vs letter grades? We only use grade points in calculations, but maybe students care about A+ vs A- on their transcript.
• Calculate some statistics.
  • What percentile do you need to be at to get a certain letter grade, per section?
    • (Maybe) This could be useful for a course-level view, and interesting (if unhelpful) for a subject-level view.
  • (QUESTION) How different are distributions for lower vs upper vs graduate?
  • For a given subject, separate statistics by term: summer vs fall/spring.
    • (QUESTION) How different are summer grades from fall/spring grades?
    • (QUESTION, maybe) Are fall and spring grades different?
• Group by instructor.
  • (Maybe) Create a model for Instructor (with a many-to-many field in Section).
    • Problem: distinct people can have the same name.
• Trend of average (or median) over time, per course or subject (or instructor; see last bullet point).
• Track enrollment over time per course or subject.
  • For a course, maybe a bar for each term, divided up by instructor.
    • (QUESTION) Does enrollment correlate with grade distribution? Let's say average or median grade. The students don't know the grade distribution in advance, but they do know the instructor.
  • For a subject, a line chart of overall enrollment should be fine. A more fine-grained display that includes classes could be overkill and really cramped, unless there's a good way to do this.
• (Bonus) Add a data import page to the admin site.
• (Bonus) It would be cool to have a GUI for playing with the API. (Who knows if anyone is going to end up using it, though...)

License

This project is covered by the MIT License in LICENSE.txt.