They boast about it being national as if that is a positive attribute. But it's exactly the opposite, because the epidemic is at different stages in different places. That alone makes it quite meaningless for anything. They noted that "being in a state with 11%–20% positive COVID-19 cases made a person nearly five times more likely to be diagnosed positive (95% CI: 1.19–21.39)."
The confidence intervals in all findings were quite wide, which shows that the data couldn't be sliced any smaller to compensate for all the possible confounders.
There was selection bias, because subjects were recruited "using a survey Web link on gaming sites, social media, customer loyalty portals, and through website intercept recruitment."
And there was no actual evidence, just subjects' say-so about their testing status.