There are three questions in Quinnipiac University’s new national poll that seem like they have some overlap.

Most Americans (67 percent) think that President Trump is not level-headed.

Most Americans (56 percent) think that Trump is not fit to serve as president.

Most Americans (69 percent) think that Trump should stop tweeting from his personal account.

This is a bit of a Rorschach test, in that you can isolate those three data points (setting aside, say, his approval rating, 36 percent, or the number of people who think he’s intelligent, 55 percent) and use them to identify the picture that you’re looking to see. But, again: They seem as if they overlap. Trump’s tweets reflect his personality, and his personality is shoot-from-the-hip, which is a nice way of saying “sometimes erratic.”

Quinnipiac’s “fitness” question — specifically: Do you think Donald Trump is fit to serve as president, or not? — is a bit broader, in that a lot of different components could play a role in someone evaluating it. Before the election, though, we saw a similar question in polling, about whether Trump was qualified for the job. Majorities consistently said no, and polling suggested that his non-levelheadedness was a big reason.

You will be unsurprised to learn that these three traits (that is: fitness, levelheadedness and tweeting) all correlate strongly. Demographic groups that strongly think Trump should stop tweeting also strongly think he’s not fit for office and so on. That’s in large part because so much of this overlaps with whether people like Trump. If you don’t like the president, you’re probably more likely to think he’s not fit for office and isn’t levelheaded.

But why use all these words when we can use a pretty and complicated graph?

The dots are how demographic groups responded to the question of Twitter use and levelheadedness vs. how people responded to the question of fitness. The dashed, colored lines are the trends those dots form.

At the center, you see a box highlighting two outlined circles. Those are the overall percentages, as articulated at the beginning of this article: Over 50 percent on each axis. In either corner, boxes highlighting the percentages for Democrats, Republicans and black respondents. The other demographic groups Quinnipiac breaks out (that is, the other dots) are four age groups, gender, education, other race/ethnic groups and whites by gender.

What I’d like you to notice is that there is only one dot under the 50 percent line on the vertical axis. That is, for only one demographic group and for only one of the questions of Trump’s tweeting or levelheadedness does less than a majority take a negative view of the president. Specifically: Fewer than half of Republicans think that Trump is not levelheaded. A majority of every other group — including working-class whites and white men — thinks that Trump should stop tweeting and isn’t levelheaded. And also notice that even a majority of Republicans think that Trump should stop tweeting from his personal account.

The question of fitness (along the horizontal axis) is less clear-cut, although there’s clearly a trend among the groups in which less than half think he’s unfit for office. Those groups are: Republicans, men, whites, white men and working-class whites. In every other group, a majority thinks Trump is unfit for office, including whites with college degrees, women and white women.

Trump isn’t going to resign from office because most Americans think he’s unfit for it. After all, he hasn’t stopped tweeting just because poll after poll suggests that Americans want him to.