Skip Navigation

Science Memes @mander.xyz

fossilesque @mander.xyz

1mo ago

sampling bias

30 comments

Technically theres always a bias of the most amusing answer so the "no" would be close to 5% at a minimum.
- Sounds like the lizardman constant
  
  That was an amazing read, thanks!
  
  Oof I didn't know that was a Scott Alexander thing
- Yep, no survey get 0.2% of any kind of response.
  
  0.2% of surveys get some kind of 0.2% response, especially if the survey is about statistics.
- It's also been shown that people often are bad a reporting accurately. So, your results may not be accurate just from that.

Survey:
I always tell the truth in surveys
I always lie in surveys
🙃
- I especially lie on corpo surveys. It's delicious hatorade.
  https://addons.mozilla.org/en-GB/firefox/addon/fake-data-haterapps/
- 2 (I don't normally but I am now >:)
- Okay, but how does this help me get past the gate?
  
  "Would the guy on the right tell me that the gate you're guarding on the left is the correct way?"
  If the answer is yes, the right gate is the correct one. If the answer is no, the left gate is the correct one.
  
  Jump magic jump
- how would the other respondent fill out this survey?

Which means 0.2% lied
- Or some portion of people generally prefer throwing surveys in the trash, but for unknown reasons decided to complete that one.
- At least 0.2% lied.
- Or they retrieved the survey from the bin, after tossing the former into the latter.

You won't get any customer complaints if you make it impossible for customers to contact you in the first place.
tapshead.jpg

0.2% margin of error.

The MNAR crowd knows what’s going on in that survey amiright?

I think it's safe to assume if they didn't respond they do not love responding to surveys.
- It's almost like that's the joke presented in comic form.
  
  It's almost like my comment was dripping with cynicism.
- Also there's at least a small set of respondents with fat fingers.
- Yeah, that's the real answer. 500 responses vs how many sent out
  
  Maybe. But if there were 10000 sent and 500 responses, that still isn't enough information. You would need to know how many were even handled and perceived by the intended recipient.
  But maybe you can work around this by canvasing in person rather than fire and forget.

30 comments