1. Understanding the arrest story with the Uniform Crime Report - NIBRS 2021 DC data

The trilogy entitled Fostering Social Impacts with Data Science is a series of data blogging works that explored all arrests made in DC in 2021 from different angles. Through data wrangling, explorative data analysis/data storytelling, and general linear regression, I found out that arrest skyrocketed in August 2021 with a rise from under 100 arrests in July to over 600 cases. Additionally, 90% of arrestees were African Americans with most of them in age 20-40. Additionally, simple assaults and intimidation accounted for around half of all arrests, and 80ish% of arrests did not involve any weapon (firearm/knife). The series is published by Towards AI on Medium.

  • Part I: Intro and Data Wrangling (Link)

  • Part II: Data Storytelling - Who and When (Link)

  • Part III: Data Storytelling II - How, Data Modeling, and Final Thoughts (Link)

$\textbf{packages}$: tidyverse, ggplot, lubridate, stringr, forcats, gt, kableExtra, sjPlot, Quarto (not a package. Quarto is a Rmarkdown-like framework used to publish high quality articles)

Available at (https://github.com/jcvincentliu/data-blogging)

  1. Nonprofit Sector In Brief Flex Dashboard and Shiny Portal:

This is the project that I worked on during my summer data science internship with Urban Institute. In this position, I developed a R Shiny portal and a flex dashboard that incorporates a handful of features to enhance user interactivity and expand user choices. Both dashboards incorporate a number of interactive charts and maps. I also created a number of easy-to- helper functions to develop a data acquisition -> data cleaning -> data visualization pipeline and reduce the running time.

$\textbf{packages}$: tidyverse, stringr, ggplotly, leaflet, httr, flex_dashboard, Shiny

Available at (https://github.com/jcvincentliu/Urban-Institute-internship-dsproject)

  1. Statistical Analysis of the Police-Public Contact Survey data:

I conducted some statistical analysis on PPCS data to examine statistical relationships between frequency of police stops and arrests and suspects' socioeconomic (age, sex, income), racial, and geo-spatial factors (where they live). I found out that these socioeconomic variables manifest strong statistical correlations with the frequencies of police interactions and arrests. Moreover, the effect doubled when these factors intertwined. My discovery shed light on the prevalence of racial profiling and neighborhood-targeted policing. I ended my analysis with recommendations on police reforms with highlights on police accountability and community policing.

$\textbf{packages}$: tiydverse, survey, base R functions (refer to a variety of base R functions, such as summary, lm, glm)

Read the full paper at (https://www.linkedin.com/in/jcvincentliu/details/featured/1635467966079/single-media-viewer/)

  1. Who were Served by Our Public Education System? Who Were Left Out?

I used the American Community Survey data, integrated by IPUMS-USA, to investigate the issue of school truancy. Through a series of analysis, I found out that as compared to Whites, racial minorities are more likely to attend public schools than private schools, and they also have a higher chance of dropping out from schools. Other socioeconomic factors including income, language fluency, immigration status, and employment, and being a Medicaid recipient or not were also highly relevant to school attendance. My findings shed light on the importance of “invisible identities”. I ended my study on the benefits of adopting culturally relevant pedagogy, replacing punitive-solutions with restorative justice, and dismantling the school-to-prison pipeline.

$\textbf{packages}$: tiydverse, base R functions (refer to a variety of base R functions, such as summary, lm, glm)