Going Beyond the Mean Effect: A Hands-on Introduction to Quantile Regression for Social Work Research

Ordinary least squares (OLS) regression remains the dominant analytic approach in social work research. However, by focusing exclusively on mean-level effects, OLS obscures how the relationship between predictors and outcomes may vary across the distribution of the outcome, a limitation that is particularly problematic in research focused on inequities, marginalized groups, or extreme risk cases. Quantile regression (QR) addresses this limitation by estimating predictor effects at different points (quantiles) of the outcome distribution, making it a powerful and underutilized tool for equity-focused research in social work.

This hands-on workshop introduces participants to the conceptual and applied use of QR to model heterogeneity in social phenomena. We will demonstrate how QR estimates conditional quantiles of an outcome (e.g., the 10th, 50th, or 90th percentiles) and allows researchers to identify whether predictors have stronger effects at the lower or upper ends of the distribution. This capability is especially valuable in understanding how structural risk factors, such as poverty, discrimination, or trauma, exert disproportionate influence on the most vulnerable subpopulations. Unlike OLS, QR does not assume homogeneity or normality in error terms and is robust to skewed outcome distributions and outliers which are common conditions in social work datasets.

Participants will receive a conceptual overview of QR, covering how it differs from traditional regression approaches, how quantiles are estimated using weighted absolute deviations, and how to interpret QR coefficients across the outcome distribution. The workshop will include guided examples using R and real-world educational data from over 170,000 kindergarten students to examine how socioeconomic disadvantage (free/reduced lunch eligibility) and vocabulary skills relate to reading success at multiple quantiles. Through these examples, we will show how QR reveals hidden disparities missed by mean-based models, such as the finding that SES has a greater negative effect on reading success among the lowest-performing children.

The workshop will also address best practices in QR, including model fit, visualization of coefficient functions, effect size interpretation, and sample size considerations for multivariate QR models. Participants will leave with a QR code template, data visualizations, and resources to extend QR to their own research.

By the end of the workshop, participants will be able to:

1. Explain how QR models heterogeneous effects across the outcome distribution

2. Identify key scenarios in social work research where QR is especially advantageous

3. Implement QR in R using the quantreg package and interpret output

4. Apply QR to examine population subgroups affected differently by structural risk factors

5. Integrate QR findings to inform targeted intervention strategies and policy recommendations

This workshop is ideal for researchers with a basic understanding of regression who are seeking innovative, equity-focused methods to better represent and serve diverse and underserved populations in social work research. Quantile regression enables social work researchers to detect how predictors differentially impact individuals across the outcome distribution, particularly those at greatest risk or with the most severe needs. By moving beyond average effects, QR supports the development of more targeted, equitable interventions and policies that better address the complex realities of marginalized populations.