Jason Williams PhD

data visualization

principles

• Visualization allows you (the researcher) or others (collaborators, readers) to understand patterns and relationships. This requires abstraction, representation, and simplification.
• Simplification means visualization is an iterative process.
• Graphic display of data allows comparison of values of one variable across levels of others--differences, trends, interactions, non-linearities. To get specific values, use a table...or a javascript display with mouse-overs, like this.
• Maximize the data:ink ratio.

current research

drug use epidemiology

• I worked with Caleb Banta-Green on transitioning from a static report in Word or PDF to a web-based Washington State opioid trends report.
• Why do different estimates of marijuana use for the same population produce different results? We compare survey sources of marijuana use estimates, contrasting sampling, mode and timing of survey, and question wording.
• Using all the information when samples taken to estimate a yearly average are censored: Statistical analysis and visualization of wastewater-based drug epidemiology data (published version available for free download until 16 October 2016)
• CSSS presentation on WBE methodology
• We have linked wastewater data on THC with the amount of THC actually sold, despite the many limitations of the state's "seed-to-sale" database. Some of the resulting frustration is expressed in this commentary in Addiction. A working paper further documenting problems and a potential solution followed a presentation covering the same territory.
• Gee whiz: The research article in Addiction has already garnered commentary and press coverage.
• The first of two brief reports using Healthy Youth Survey data covers the relationship between simple poverty indicators--a food insecurity screener or free/reduced price lunch program participation--and cannabis use in Children and Youth Services Review.

drug interventions

• Evaluation of a treatment decision-making model to promote opioid use disorder treatment initiation, published in Drug and Alcohol Dependence.
• Evaluation of the first iteration of the Medications First model of low barrier buprenorphine in Washington, in Substance Abuse and Rehabilitation.
• Under the title The Promise and Peril of Large Administrative Datasets, I presented on lessons learned from this project at the Center for Statistics and the Social Sciences seminar in May 2024.

child welfare

• Predictive analytics is probably a good thing, but it requires good data on the front end and a plan and required resources to do something about the prediction results on the other end. In our paper in Child Abuse and Neglect, the resulting quality assurance procedures may have resulted in better investigations...but that is not the outcome we evaluated.

Recent Updates

[09/16/24] Meds-First eval published

[01/08/24] HYS brief report published

[09/19/22] New papers

[06/09/21] Updated sections, new publications

[10/20/20] On-going and largely unadorned COVID-19 visualization

[03/04/20] Added CSSS presentation on wastewater-based epidemiology technical and statistical issues

[12/20/19] Free access (until February) to Drug and Alcohol Dependence article on our evaluation of a Treatment Decision Making model for people leaving prison.

[08/06/19] "Gee whiz"

[08/20/18] Estimating cannabinoids sold in noisy sales data

[11/06/17] Marijuana data commentary in Addiction

[07/11/17] First HIV treatment paper out in AIDS and Behavior

[06/24/16] Added 2 more papers, published in Science of the Total Environment and the American Indian Law Journal

[04/14/16] Updated JSSWR publication, added WCPC presentation

[08/05/15] Added 2 papers published this summer plus ADAI work

[08/04/15] Looks like the whole dissertation is available at ProQuest now

[03/11/14] Received notice from ProQuest about dissertation publication, added link to PQ page

[06/07/13] Added a first stab at Tableau using dissertation data

[11/10/12] Added working papers, research examples, shameless plug for foster care alumni study book

Past research

A NEW LOOK AT POVERTY: INCORPORATING TIME INTO CONCEPTIONS OF POVERTY

• Economics is the study of how actors allocate scarce resources to produce "utility" in the face of constraints. Time, like income, is a scarce resource we use to produce utility (or, if you prefer, well-being). In this paper, we argue for considering time alongside income in defining poverty, review past attempts to measure time poverty, and highlight how measurement of (time) poverty can improve. accepted manuscript version; the final publication is available at Springer via http://dx.doi.org/10.1007/s11205-015-1029-z and now from Springer Nature SharedIt.
• The next step: Exploring the topography of well-being over time and income, here in a presentation at the West Coast Poverty Center, and here in an article in the Journal of Happiness Studies.

childhood income and poverty dynamics and adult adiposity

• A growing literature is connecting childhood SES conditions to adult obesity. My dissertation (abstract available via ProQuest) built on such life course investigations by examining different definitions of poverty and of income instability, and the extent and timing of these conditions, in influencing later BMI.
• From my NLSY data, here are the basic relationships between childhood income (using the NLSY Total Net Family Income, with mother's reported income characterized in terms of the year the child was born, turned 1, etc.) and adult BMI--or at least as Tableau views them.
• US adiposity dynamics: An analysis of adult BMI, percent body fat, and obesity in recent waves of NHANES. This project was for my survey sampling class, and uses the two bits of R code for survey analysis below. It parallels Burkhauser, Cawley, & Schmeiser (2009) in using the skinfold thickness measures in NHANES to create an alternative definition of obesity.

Travels with NLSY (no longer updating)

• My attempt to capture what I've learned about dealing with the NLSY so that other students don't have to reinvent the wheel.

writing

obesity policy

• Taxing sugar: An analysis of how science did not influence policy. A revised version of a policy process piece I did for a seminar on interactions between research and policy making.

drug use and HIV treatment

• I worked with Bryan Hartzler at ADAI on a large dataset comprised of administrative records from multiple HIV treatment sites. One article, on how substance use disorders influence treatment compliance, is now available in AIDS and Behavior.

foster care

• Compliance with the Indian Child Welfare Act: In which we discuss the background of ICWA, potentially measurable compliance points, basic social science measurement principles, and the difficulties of measuring ICWA compliance. Ironically, about a week after publication, these difficulties decreased slightly with the publication of official rules from the Bureau of Indian Affairs on implementation of the 1978 Act.
• An analysis of factors related to staff endorsement of family group conferencing: A Bayesian Model Averaging look at characteristics that may predict staff buy-in and referral to FGC, from my (first) return to Casey Family Programs. Published Sep 2015 in Journal of the Society for Social Work Research.
• Predicting placement disruption: An event history analysis of factors related to the disruption of a family foster care placement.
• The foster care alumni study: Data for the above paper came from a project I worked on from 1999 to around 2008, culminating in this book, now in a Kindle edition!

school violence

• A policy process study of macropolitical attention to school violence and how it gets distorted by dramatic events (Believe it or not, but mass shootings were once relatively rare events.)

r code

survey analysis, based on LUMLEY'S SURVEY PACKAGE

• non-survey equivalent of svysmooth: This function estimates kernel densities for unweighted data using the exact same methods as svysmooth's default behavior.
• survey-weighted komogorov-smirnov test: This function allows conducting a kolmogorov-smirnov test of whether two distributions could have arisen from the same distribution with either two survey design objects, two variables within the same survey object, or one survey object and one vector of unweighted observations.

visualizing data

• activity time tablechart
• R code for the tablechart