> So you are projecting opinions of 0.0001% of Trump supporters (or trolls, who knows) to 60M+ people who voted for him?
1. No, I'm looking at differences in properties of samples drawn from different populations to estimate differences in those populations. Trump-related online forums have a higher fraction of racists than Clinton, Sanders, Cruz, Rubio, Bush, and Kasich forums do.
If Trump supporters in general had the same level of racism as supporters of those other candidates, then we would expect Trump forums to show about the same level of racism as the forums of those other candidates.
2. You seem to think the size of a sample as a percentage of the population is important. It is not. For large populations it is the absolute number of people in the sample that is important, not the percentage of the population.
I am making the assumption that supporters of various candidates are all about equally likely to decide to post on their candidate's forums. It is possible that there is something about Trump supporters that makes racist Trump supporters more prone to make racist forum posts.
That is what you should be focusing on to criticize my inferences about Trump supporters, rather than basing it on your inadequate understanding of elementary probability and how it applies to determining the sample size needed to make reasonably inferences about a population from a sample.
> I see a bright career at WaPo or NYT for you :-)
Have an upvote. I like it when people who are wrong are at least feisty about it.