Read on Twitter
Speaking as someone who researches public understanding of statistics (and survey methodology), I am begging you: Please, please, please stop sharing the survey that "2 out of 5 Americans plan to attend a large family gathering." 
I would not be surprised whatsoever to learn that that a shockingly large number of people plan to attend irresponsibly large gatherings with no or minimal mitigation - but this survey can't tell you that.
2/
2/
As best as I can tell from the topline, respondents were asked how likely they were to "invite people to or attend a gathering with no more than 10 close family members.” 38% said they were not very likely or not at all likely to do so.
3/
3/
This is ASTONISHINGLY poor question wording for multiple reasons. First of all, negative phrasing in this type of question leads to awkward double-negatives that can make the responses hard to interpret.
4/
4/
(What does "I am not likely to attend a gathering with no more than 10 close family members" mean? Honestly, I find it this deeply ambiguous - and ambiguous survey responses lead to confusing statistics.)
5/
5/
Multiple clauses in this type of question also make the responses hard to interpret. They're called "double-barreled questions" and it's hard to know if people are paying attention to just one part or to the whole thing.
6/
6/
(To make this concrete, imagine you're spending the holiday at home alone with your cat and your one roommate. You're not at all likely to invite people to or attend *any gathering at all* ...
7/
7/
... so is the answer "not at all likely" since you're not gathering? Or is it "extremely likely" since you're definitely avoiding gatherings of 10+ ?)
8/
8/
Assuming the topline accurately represents the order in which the questions were asked, respondents were asked this question immediately after being asked if they would stay within their household. But people interpret questions in context.
9/
9/
If a respondent has literally just informed the pollster or online survey that they are definitely staying within their immediate family, they're probably going to be more likely to focus on the "invite or attend" rather than the "no more than 10 people" part.
10/
10/
tl;dr: this @BadSurveyQ means that the 38% number includes BOTH people who are throwing a big-ass Thanksgiving party AND at least some people who are not leaving their own damn homes. That makes it pretty useless in figuring out how many people are having big parties.
11/
11/
And by the way, all of these issues could have been avoided by asking people how likely they were to "avoid gatherings of 10 people or more."
12/12
12/12
P.S. This is why survey designers need linguists.
P.P.S. *Even if* the survey said what they think it says, you should always read headlines "X% of Americans do Y" as "Most likely, (X+/-4)% of Americans do Y." But that is a whole other thread on sampling and margin of error.
