4 ms·

Correlated: Discover surprising correlations

21 points by jawns 14y ago

tfb 14y ago
I really hope it becomes standard for websites like this one to no longer require a full signup to use it. Maybe use a combination of sessions, IP addresses, browser fingerprints, etc. to differentiate anonymous users. I understand that the creator(s) of the site want to gather as much contact information as possible, but those who care to be a part of the community will do just that, while the rest (like myself) just want to try it out immediately... without filling out another sign up form. I submitted my answer to the corn on the cob question but didn't make it any further than that because it immediately asked me to fill in a handful of fields to sign up. If I was able to continue without having to do that, I would have probably stuck around longer to provide the site with answers to other questions (providing value) and possibly even gotten sucked in and if I saw that the site was well-made, become a long-lasting member of the site's community.
Xcelerate 14y ago
Holy cow. I hope that shower statistic isn't really true!
kvnn 14y ago
Is it safe to take a survey of 50 people seriously?
- jrockway 14y ago
  No. This site should be called unknown-quantity-of-noise-on-top-of-unknown-quality-of-signal.org. But to be fair, that is a lot of typing.
  - twiceaday 14y ago
    This wont stop the vast majority of people from quoting this site as if their noise = correlation = causation. You will never go broke underestimating the intelligence of the public.
jrockway 14y ago
It's important to change the expression "X% of people" to "X% of people browsing the Internet with nothing better to do than answer silly questions". For example: In general, 61 percent of people who search for a blog on correlations and submit answers to poll questions there claim to pee in the shower. But among those people who have nothing better to do than click buttons on blogs and who claim they would use a bidet if available, 79 percent claim they pee in the shower. With some level of rigor you can correct for self-selection and lying, but I doubt they are doing that here.
goodside 14y ago
'Each day, a new survey is posted — something like "Do you like math?" or "Have you ever gone on a blind date?" 'At the end of the day, the results of the survey are compared with the results of all previous surveys, and the two outcomes with the strongest link are highlighted.' Insert face in palm. No. You cannot do this. Nearly all of the correlations found on this blog are pure noise. This is the data-science equivalent of staring at a cloud, noticing it looks like a balloon dog, and thinking "Jesus, why would anyone go to the trouble of making a cloud that looks like a balloon dog? Think of the engineering effort involved! Why isn't this in the news?" Google for "multiple comparison issues" and "Bonferroni correction" if you want more technical detail on how stupid this is.
- 1337biz 14y ago
  Can't upvote this comment hard enough. It is sad that these correlation=causation myths are getting promoted in such a superficial form. And the media is once again eager to jump on the bandwagon Perigee, an imprint of Penguin Books, to produce a book based on Correlated. The book will feature some of the stats from the blog, as well as a bunch of never-before-published correlations. His blog on the other hand is actually quite interesting to read http://coding.pressbin.com/88/How-my-new-site-Correlated-works/ http://coding.pressbin.com/88/How-my-new-site-Correlated-wor...
SeanDav 14y ago
28% of HN people looking at the OP site will regret wasting 2 minutes of their life.
jawns 14y ago
Hi, I'm the OP. I'd just like to encourage the critics in this thread to read "A response to Correlated's critics," which is linked to from the homepage: http://coding.pressbin.com/108/A-response-to-Correlateds-critics/ http://coding.pressbin.com/108/A-response-to-Correlateds-cri...
jstanley 14y ago
Everyone hating on this: lighten up. I am pretty sure they know that these correlations are nonsense. It is still a funny site.

21 points by jawns 14y ago

tfb 14y ago

I really hope it becomes standard for websites like this one to no longer require a full signup to use it. Maybe use a combination of sessions, IP addresses, browser fingerprints, etc. to differentiate anonymous users. I understand that the creator(s) of the site want to gather as much contact information as possible, but those who care to be a part of the community will do just that, while the rest (like myself) just want to try it out immediately... without filling out another sign up form. I submitted my answer to the corn on the cob question but didn't make it any further than that because it immediately asked me to fill in a handful of fields to sign up. If I was able to continue without having to do that, I would have probably stuck around longer to provide the site with answers to other questions (providing value) and possibly even gotten sucked in and if I saw that the site was well-made, become a long-lasting member of the site's community.

Xcelerate 14y ago

Holy cow. I hope that shower statistic isn't really true!

kvnn 14y ago

Is it safe to take a survey of 50 people seriously?

jrockway 14y ago
No. This site should be called unknown-quantity-of-noise-on-top-of-unknown-quality-of-signal.org. But to be fair, that is a lot of typing.
- twiceaday 14y ago
  This wont stop the vast majority of people from quoting this site as if their noise = correlation = causation. You will never go broke underestimating the intelligence of the public.

jrockway 14y ago

It's important to change the expression "X% of people" to "X% of people browsing the Internet with nothing better to do than answer silly questions". For example: In general, 61 percent of people who search for a blog on correlations and submit answers to poll questions there claim to pee in the shower. But among those people who have nothing better to do than click buttons on blogs and who claim they would use a bidet if available, 79 percent claim they pee in the shower. With some level of rigor you can correct for self-selection and lying, but I doubt they are doing that here.

goodside 14y ago

'Each day, a new survey is posted — something like "Do you like math?" or "Have you ever gone on a blind date?" 'At the end of the day, the results of the survey are compared with the results of all previous surveys, and the two outcomes with the strongest link are highlighted.' Insert face in palm. No. You cannot do this. Nearly all of the correlations found on this blog are pure noise. This is the data-science equivalent of staring at a cloud, noticing it looks like a balloon dog, and thinking "Jesus, why would anyone go to the trouble of making a cloud that looks like a balloon dog? Think of the engineering effort involved! Why isn't this in the news?" Google for "multiple comparison issues" and "Bonferroni correction" if you want more technical detail on how stupid this is.

1337biz 14y ago
Can't upvote this comment hard enough. It is sad that these correlation=causation myths are getting promoted in such a superficial form. And the media is once again eager to jump on the bandwagon Perigee, an imprint of Penguin Books, to produce a book based on Correlated. The book will feature some of the stats from the blog, as well as a bunch of never-before-published correlations. His blog on the other hand is actually quite interesting to read http://coding.pressbin.com/88/How-my-new-site-Correlated-works/ http://coding.pressbin.com/88/How-my-new-site-Correlated-wor...

SeanDav 14y ago

28% of HN people looking at the OP site will regret wasting 2 minutes of their life.

jawns 14y ago

Hi, I'm the OP. I'd just like to encourage the critics in this thread to read "A response to Correlated's critics," which is linked to from the homepage: http://coding.pressbin.com/108/A-response-to-Correlateds-critics/ http://coding.pressbin.com/108/A-response-to-Correlateds-cri...

jstanley 14y ago

Everyone hating on this: lighten up. I am pretty sure they know that these correlations are nonsense. It is still a funny site.