jenwhitson/sentencing_project

Capstone for Nashville Software School Data Analytics Bootcamp. Examines racial disparities in sentencing from 2015 to 2019 in Tennessee. Technologies used: Python, SAS, Excel, PowerPoint

sentencing_project's Issues

Plot census demos v. prison demos

Already got prison demos plotted, just need to put side by side with state stuff. May need to convert stuff to percentages.

Add additional dimensions noted after review of opafy dimensions table and rerun queries in SAS to add to csv files

Review sentencing guidelines on USSC.gov

Calculate state demographic percentages 2015-2019 using census data.

race, gender, age, education level if possible. will need to be done in sql, as census reporter files are stored as postgres backup.
4/15/21 - can't get backup to load. mahesh is working on this. in the interim, we will use data directly from the census reporter from 2019.

Look at sentencing enhancement categories and calculation categories, then make plan for how to analyze sentencing length using this data.

Make histo of age at sentencing

According to Rachel Elise Barkow, most people age out of violent crime by 40. Exceptions are fraud, gambling and familial child molestation.

pull 2015-2019 census data from census reporter

Look at breakdown of DISPOSIT column grouped by MONRACE, MONSEX, and HISPORIG. This will tell us what percentage of each ethnicity and gender plead vs what percentage go to trial.

subset totprisn for pleas v. non-pleas

Find out if we can figure out how many people were arrested and never sentenced.

May be able to compare with arrest records somehow, but I'm not sure if it will be possible. I think it would be instructive to know what percentage of people are arrested but never charged. Those people have to interact with law enforcement, but are presumably innocent if never charged. I think the amount of interaction with law enforcement for innocent people is another interesting dimension.

Add talking points to presentation

Review sentencing table on USSC.gov and take notes

Find out if this data is only for people charged with federal offenses.

Make grouped bar chart of percent general pop v. percent prison sentences.

Create clipped sentence column that stops at 720

Go through python notebook and export any useful charts

Doesn't have to be ones we'll use, can put them in the readme or somewhere else.

Stopped 4/21/21 on after the plea v. trial chart.

Make cards for percentage more likely someone is to be sentenced than other groups.

Change savefig settings to 300 dpi instead of 1000.

Explore MNTH and PCNTDEPT to see how they play together and if any info can be gleaned from them. And how they play with RANGEPT.

Add new dimensions to opafy dimension table.

Subset people who served no time and see what we can learn about them.

May want to examine their pleas, provision of substantial assistance, etc.

Add immigration stats to presentation

Create dimension table for opafy tables and categorize each dimension.

Subset longest sentences and explore.

Think over 10 years is a good dividing line.

Merge 2015-2019 TN tables if possible

Need to figure out how to deal with missing columns first.

Run value count for MONRACE, MONSEX, and HISPORIG.

Determine which districts are in TN and subset opafy tables for districts located in TN

Examine pre-existing statistical dimensions of opafy tables grouped by race

Brainstorm ways to sanity check existing tables and set tasks for any ways identified.

Calculate average fines by race and crime type.

Determine which columns need to be captured and query opafy files in SAS with just these columns. Save as csv's.

Create categories for sentence lengths (maybe <1 year, 1-3 years, 3-5 years, 5-10 years, 10-20 years, 20+ years) and then create grouped bar chart by race.

Look at pretrial detention rates by race/gender

Look at average sentences by crime types

This is basically done, but don't feel like I have a good sense of what the data is telling me. Why do hispanics have such higher sentences for going to trial when there's basically nothing showing up as excessive in the crime type breakdown? Immigration crimes look pretty low.

Figure out exactly which columns are missing from each year.