Hacker News Reader

Comment by mirashii | Hacker News Reader

It's a bit concerning that this page doesn't really reference the order of magnitude of data that they're working with, statistical significance, or anything to indicate that the difference between the two percentages they're comparing is something that's actually worth comparing, and not something that can or is likely to be explained away due to small sample size.

View on HN

That's a fair concern. The sample sizes here are relatively high: you can probably back some estimates out yourself from the numbers given in the post (millions of dollars raised, mean donation of $88, 14% of donations in the last week were repeat, etc). I'll also say that the comparisons in the post are significant to a 99% confidence level.

Also there are many other variables that could affect this analysis that aren't explored or mentioned. Just two I can think of off the top of my head:

- Users logged into stripe probably buy more stuff online. Maybe they donate more money because they have more money.

- The fact that a ton of repeat donations came in the last week might have been due to email campaigns or some other external thing. It isn't necessarily a fundamental principle of fundraising drives (though it certainly could be). Same thing could explain mobile usage driving up (SMS campaign or something).

Overall I think it is cool that data is being shared like this, and it is interesting. But at the same time, the post shouldn't be speaking about data in such absolutes... no confidence intervals, few caveats mentioned or explored, conclusions are made with bolded certainty.

From the downloadable dataset of donations, the current total amount of donations is 62,769 donations with 50,802 unique donators, which is a healthy amount.

However, that doesn't indicate how many users fell into the "had a Stripe account already" bucket. (I'd wager not many.)

(Off topic aside about data integrity: the downloadable data, from October onward, has donation dates from 2018.)

This is Mario from Mayday PAC. Yeah, our last release had those 2018 issues. The data actually comes from several different vendors and we plan on normalizing it all better post election. The 2018 data should have been registered as 2014. (Obviously not 2018)

Reply on news.ycombinator.com