Metrics and Statistical Analyses of JEDI-related Data

Studying racial or ethnic disparities in medical research often involves using person-level race or ethnicity to define group membership and subsequently estimating regression-adjusted absolute or relative measures of between-group differences in a health outcome. However, research that uses individual-level race/ethnicity as a variable without considering racial segregation (or diversity) of study participant's residential neighborhoods generates biased inferences about racial/ethnic disparities because segregation confounds the role of individual-level race/ethnicity. As a result, such research incorrectly attributes racial/ethnic disparities to person-level characteristics or behaviors instead of community-level factors and their upstream determinants including structural racism. We will discuss the application of various statistical designs and analyses that can be deployed to account for residential segregation. Using an empirical example, we also show results that account for residential segregation (or diversity) tend to be robust to the choice of segregation or racial/ethnic diversity used.  

The US Census Bureau's Diversity Index is based on the likelihood that two randomly selected individuals in a given geography are from different race and ethnic groups. These data are available at the state and county level for 2010 and 2020. This presentation introduces the US Census Diversity Index and examples of analysis using it. An important limitation in the available data is illustrated in a case study from the Detroit Metropolitan area. The Diversity Index for Wayne County, Michigan, which includes the city of Detroit, indicates a high level of racial diversity. However, applying the methodology of the US Census Diversity Index to the city of Detroit and certain inner suburbs in comparison to the remainder of Wayne County, the Detroit area is found to be one of the least diverse and most racially segregated areas in the United States. In this way, applying the methodology of the US Census Diversity Index to metropolitan areas allows for direct analysis of implicit, de facto segregation. An outline is provided to assist local ASA chapters to host a hackathon to use the Diversity Index to analyze racial and ethnic diversity in their area.
 

Many agencies have pursued interests regarding increasing diversity and inclusion in varied spaces. How can one quantitatively assess effectiveness in these areas? What statistical measures or approaches exist for addressing diversity and inclusion? This talk offers a review of existing proposed metrics offered across various disciplines, serving as a building block for relevant research.