Events

Events

Oct 05 2022

Should surveys produce more contextual features? Comparing contextual features by alternative definitions of neighborhoods

Shiyu Zhang, Michigan Program in Survey and Data Science

When:
October 5, 2022
12:00 – 1:00 pm

Join via Zoom:
Zoom Link
Meeting ID: 99290637991
Meeting Password: 1949

Abstract:
Should surveys produce more contextual features? Comparing contextual features by alternative definitions of neighborhoods.

Shiyu Zhang is a PhD candidate at the Michigan Program in Survey and Data Science. Before arriving at Michigan, she received master’s degrees in immigration study, sociology and data science, and a bachelor’s degree in psychology. Shiyu’s dissertation focuses on the effect of adaptive survey design on estimates. She is also interested in collecting and using neighborhood features as auxiliary variables.

An important methodological challenge in studying neighborhood effects is how to geographically define “neighborhoods” and create contextual features to characterize the areas. In quantitative research that uses survey data, contextual features are commonly defined by census geographies like census tracts and block groups. However, the literature has called for expanding the definition of neighborhoods beyond census boundaries and exploring contextual features in geographic areas more relevant to the studied individuals.

In this research, we compare social and built environment features of neighborhoods based on three geographic definitions (i.e., census tracts, residential buffers, and respondent-informed neighborhoods). We evaluate how the alternatively defined measures influence the detected associations between contextual features and health outcomes. Our findings suggest that the neighborhood definition matters. Therefore, other than simply offering linkages to census boundaries based on participants’ geocoded location, surveys may enrich the data and support further research by producing and releasing case-specific contextual features.