Fig. 1

Regression where the outcome variable was the proportion of each document dedicated to each topic. Values were generated from a regression where the outcome variable was the proportion of each document dedicated to each topic, given the selected STM model, and with various vaccine events as main explanatory variables. Topics on the right of the zero line were more likely to be brought up after the indicated event. Confidence intervals (95%) included both regression uncertainty and measurement uncertainty from the STM model. A, B, and C refer to the above indicated events