From Electoral-Vote.com:
A theory has been circulating that the Donald Trump tweets that come from an Android device are from the candidate himself, while the ones that come from an iPhone are the work of his staff. David Robinson, a data scientist who works for Stack Overflow, decided to test the theory. His conclusion: It's absolutely correct.
Robinson did some text-mining (using R) to analyze roughly 1,400 tweets from Trump's timeline, and demonstrated conclusively that the iPhone tweets are substantively different than the Android tweets. The former tend to come later at night, and are vastly more likely to incorporate hashtags, images, and links. The latter tend to come in the morning, and are much more likely to be copied and pasted from other people's tweets. In terms of word choice, the iPhone tweets tend to be more neutral, with their three most-used phrases being "join," "#trump2016," and "#makeamericagreatagain." The Android tweets tend to be more emotionally charged, with their three most-used phrases being "badly," "crazy," and "weak."
(Score: 0) by Anonymous Coward on Monday August 15 2016, @03:17AM
The explanation in TFA is pretty straightforward; I don't think R had much to do with it. It doesn't mean you have to agree with his conclusions, since it's not a traditional statistical analysis, though most looked pretty convincing.
Some might quibble over his decision to exclude tweets with quotes, stuff like that.