Fig 1.
Conceptual diagrams for mediation analysis.
cEM means the variance of the error term of mediators; cEO means the variance of the error term of outcomes; cC means the variance of controls; cP means the variance of predictors; rCP refers to the covariance matrix between control variables and predictors.
Table 1.
Summary of county-level variables.
Fig 2.
Spatial distribution of human mobility and COVID-19 health outcomes in the contiguous US.
Each panel represents one metric from (a) average staying home (%) from 1 March to 31 December 2020 compared to 2019 baseline, (b) average visit change (%) from 1 March to 31 December 2020, (c) cumulative cases/100,000 by the end of 2020, and (d) cumulative deaths/100 cases by the end of 2020.
Fig 3.
Temporal evolution of human mobility and COVID outcomes.
Each row depicts a metric from (a) weekly percentage change in POI visits compared to 2019 baseline, (b) weekly percentage of residents staying home, (c) weekly new confirmed cases/100,000, and (d) weekly new deaths/100 cases. Each column represents a racial group, with each curve denoting one quintile. Sample comprises 3,108 contiguous US counties. Y-axis limits are shared by each row individually.
Table 2.
T-tests of difference in mean between quintiles of racial compositions (county level).
Table 3.
Summary of SEM goodness-of-fit.
Table 4.
Output of structural equation models (mediator: Staying home (%)).
Fig 4.
To clearly show the most important structure, we only depicted the paths with 1) P-value < 0.05, 2) standardized direct effect greater than 0.1, or 3) standardized covariate greater than 0.3. Double-headed dotted arrows represent the covariates between exogenous variables. Solid arrows represent the direct effect. The width of the arrow is proportional to the magnitude of standardized path coefficients. Red refers to a positive estimation while green represents a negative one. Similar diagrams can be found in S2 Fig with visit change (%) as mediator.
Table 5.
Output of mediation effects (mediator: Staying home (%)).