Table 3.

Regression models evaluating the relationship between Twitter sentiments and emotions and HPS^a vaccine hesitancy and compliance.

Model	Features	Outcome
Model 1a (GLMb)	Percentage of positive sentiment Percentage of joy Percentage of surprise Percentage of trust Percentage of anticipation Percentage of survey week (fixed effect) Percentage of metropolitan area (fixed effect)	Percentage of vaccine-compliant HPS respondents
Model 1b (GLMMc)	Percentage of positive sentiment Percentage of joy Percentage of surprise Percentage of trust Percentage of anticipation Percentage of percentage of survey week (fixed effect) Percentage of metropolitan area (random effect)	Percentage of vaccine-compliant HPS respondents
Model 2a (GLM)	Percentage of negative sentiment Percentage of anger Percentage of disgust Percentage of sadness Percentage of fear Percentage of anticipation Survey week (fixed effect) Metropolitan area (fixed effect)	Percentage of vaccine-hesitant HPS respondents
Model 2b (GLMM)	Percentage of negative sentiment Percentage of anger Percentage of disgust Percentage of sadness Percentage of fear Percentage of anticipation Survey week (fixed effect) Metropolitan area (random effect)	Percentage of vaccine-hesitant HPS respondents

^aHPS: Household Pulse Survey.

^bGLM: generalized linear model.

^cGLMM: generalized linear mixed model.