4 ms·

>Instead of thinking "9/10 people dislike this UI", it's probably closer to "9 people out of all Firefox users dislike this UI". Yeah, even with your Bayes men

>Instead of thinking "9/10 people dislike this UI", it's probably closer to "9 people out of all Firefox users dislike this UI". Yeah, even with your Bayes mention I'm gonna go ahead and get a second opinion on your statistical analysis there. I know, there's always someone in every crowd who quibbles with a poster's perspective, but you're hanging your entire sentiment on this one.

cdrini 5y ago
I'm not following your comment; can you expand? To rephrase what I was saying, I think that when we read 10 comments, of which nine are negative, we think "90% of people must dislike this". I don't think most people would actually think 90%, but I do think we "get a feeling" it's a very high percentage. We get this impression that the 10 comments are by 10 random people, so it's close to a random sample. But it's not, it's a very biased selection. The very post itself acts as a selection filter, only selecting people with certain properties. In this case, people with strong opinions about Firefox's UI. In short, the denominator isn't 10, but something much much greater. And this isn't a stastical analysis (not yet anyways!), just a statistics/psychology hunch :)
- rhizome 5y ago
  >I think that when we read 10 comments, of which nine are negative, we think "90% of people must dislike this" I don't. When I see something like that with such a small sample size, I think that the downvoters just got there first. That could be because the YouTube algorithm put the video in front of people who happen to have higher standards, or who are bitchier, or who are familiar with the poster and their disruptive complaints over on the Slack for $X. Could be anything...anything other than generalizable overall quality.
  - cdrini 5y ago
    I'm hypothesising that this is a bias people have when reading posts in the internet. People are susceptible to it to different degrees. And your statement does show this bias in a different way. You comment on the 10 being too small a sample size. That implies that if there were more posts, say 90 vs 10, you would draw some conclusions about the question "Do people like the new UI?". In the same way you're saying that the sample would be biased because someone posted on YouTube, I'm saying it will _always_ be biased, regardless of size, because the post itself (and all posts) attracts a biased sample. A larger sample is good for a random sample, but I think posts attract a sufficiently biased sample by their very nature, so increasing the sample size won't help.

cdrini 5y ago

I'm not following your comment; can you expand? To rephrase what I was saying, I think that when we read 10 comments, of which nine are negative, we think "90% of people must dislike this". I don't think most people would actually think 90%, but I do think we "get a feeling" it's a very high percentage. We get this impression that the 10 comments are by 10 random people, so it's close to a random sample. But it's not, it's a very biased selection. The very post itself acts as a selection filter, only selecting people with certain properties. In this case, people with strong opinions about Firefox's UI. In short, the denominator isn't 10, but something much much greater. And this isn't a stastical analysis (not yet anyways!), just a statistics/psychology hunch :)

rhizome 5y ago
>I think that when we read 10 comments, of which nine are negative, we think "90% of people must dislike this" I don't. When I see something like that with such a small sample size, I think that the downvoters just got there first. That could be because the YouTube algorithm put the video in front of people who happen to have higher standards, or who are bitchier, or who are familiar with the poster and their disruptive complaints over on the Slack for $X. Could be anything...anything other than generalizable overall quality.
- cdrini 5y ago
  I'm hypothesising that this is a bias people have when reading posts in the internet. People are susceptible to it to different degrees. And your statement does show this bias in a different way. You comment on the 10 being too small a sample size. That implies that if there were more posts, say 90 vs 10, you would draw some conclusions about the question "Do people like the new UI?". In the same way you're saying that the sample would be biased because someone posted on YouTube, I'm saying it will _always_ be biased, regardless of size, because the post itself (and all posts) attracts a biased sample. A larger sample is good for a random sample, but I think posts attract a sufficiently biased sample by their very nature, so increasing the sample size won't help.