We re-estimated both models in samples excluding participants who were never subject to a “shelter-in place” order, and in the sample that were never subject to an order, or who had the order lifted at some stage during the study.