Muse: OFF TOPIC: This Is What Your Brain Looks Like On Evidence...
37 Responses
First ←Older Page 1 2 Newer→ Last
-
Tim Hannah, in reply to
Methinks it’s time for Messers Key and Goff to start hearing from some families (and registered electors) that Bob most definitely does not speak for
Read Bob McCoskrie in the Dompost yesterday. Made me want to get a divorce.
Bob, if you don't piss off my children might* be brought up by divorced parents. Bob, think of the children.
*Haven't actually run this past all interested parties, probably should first, though I'm sure she'd understand.
-
bmk, in reply to
Heh. Or ring up a few hundred people randomly and just ask the question. Polling isn't rocket science
Well it isn't rocket science but you should also record demographics of the person and weight accordingly, you should also try to randomise the person being asked (otherwise this introduces bias as some people are more likely than others to answer the phone) which is usually being done by asking for the person in the household who will have their birthday next to answer the questions.
So while not rocket science it takes a bit of care and work to get reasonable results - you should also try to get at least 500 answers which would probably involve ringing at least 5,000 numbers (pure guess as to response rate - but I can't imagine getting more than one in ten).
Polling companies to reduce costs often combine several polls on the one person which makes the process more economic.
-
BenWilson, in reply to
Polling companies to reduce costs often combine several polls on the one person which makes the process more economic.
Or they've already done the poll and can just sell you the answer. But sure, they have economies of scale, that's why people use them. And ringing people up is very dull work (for me at least). But it sometimes does help when people want answers to survey questions to remember that all they're really doing is phoning people up and asking them questions. It all depends how much money is worth to you, and for that matter, how much accuracy you really want. If it's "eight out of ten people surveyed said they thought gay marriage was already in law" then you might get a good enough answer to that out of calling up 30 people. If you want it down to 0.5% accurate, and to make whizzy charts out of it, and present it to a board of executives, then you'd spend the money, without question.
-
bmk, in reply to
If it's "eight out of ten people surveyed said they thought gay marriage was already in law" then you might get a good enough answer to that out of calling up 30 people.
Problem is with thirty people you'd have a margin of error of around 18% if the sample were perfect. So that would mean 'somewhere between 6 and 10 out of 10 people thought gay marriage was already in law' and in reality the gap would be much higher as the odds are there would be biases introduced through just ringing up 30 people without taking care to weight for demographics and randomising the answer person. So a genuine answer would be something like 'between 3 and 10 out of 10 people thought gay marriage was already in law'.
A too small sample is practically useless. I remember a scientist once telling me the most common mistakes people make are incorrect cause-co-relation assumptions and making assumptions based on an insufficient sample size. This is really born out when you talk to people and there reason for believing in something is an anecdote they heard from a friend of a friend and to their mind this is conclusive proof.
-
BenWilson, in reply to
Yup, stats is great, but sometimes you aren't actually after that much accuracy in your data. You might just be trying to decide if it's even worth further investigation.
-
How did politicians ever do anything before the rise of pollsters and focus gropes? (ETA: That a typo to keep.) Jus’ asking.
-
I understand that Research New Zealand will be polling on the issue next month.
-
Kumara Republic, in reply to
How did politicians ever do anything before the rise of pollsters and focus gropes? (ETA: That a typo to keep.) Jus’ asking.
For a moment there I almost read 'focus groupies'. Still, it probably wouldn't be too distantly related.
A wider issue is that polling companies appear not to have discovered cellphones yet. I wonder how polling is done in the home of Nokia and Linux?
-
bmk, in reply to
A wider issue is that polling companies appear not to have discovered cellphones yet.
Apparently as yet this hasn't been a significant problem as they have found the people who don't have land-line phones vote similarly to the people who do have land-lines and match their demographic. So by weighting the votes of that demographic who do have land-lines more heavily to compensate for the missed people (those with cell phones only) they end up with a fairly accurate result.
-
Idiot Savant, in reply to
Polling. I want some fucking polling on this.
ResearchNZ poll just out today: 60 - 34% in favour. Question was "should same sex couples also be allowed to marry?"
-
Emma Hart, in reply to
ResearchNZ poll just out today: 60 - 34% in favour. Question was "should same sex couples also be allowed to marry?"
That's fabulous. Also, about what I'd expect as it's in line with Australia. Soon as that goes up online I'll be all over it.
-
Idiot Savant, in reply to
Email me, I'll send it to you.
Post your response…
This topic is closed.