Into the Night
Verified User
Polling isn't a science. Science is a set of falsifiable theories.what I see is polling science getting more and more accurate over the last 70 years
It often is.Polling isn't make believe
The raw data is very rarely published. If it was, the media couldn't spin it.The issue is the media will take raw data and spin it how they want.
They are wrong on their own. They are not science. They don't even comply with the requirements of the rules of statistical mathematics.Like for example how I always talk about the difference of sampling "american adults" vs. "registered voters" vs. "likely voters"
None of those polls would be wrong on their own,
No, there is no set. Raw data is not published.the problem is when people try to extrapolate from one set of data and pretend it's a different set.
Irrelevant. Polls are incapable of prediction the future. So is statistical mathematics.So sampling "adult americans" to demonstrate something about electoral chances is stupid, because if you want to know how an election is going to go you need to poll people that will vote, not randos that are staying at home.
This creates a math error.Now lets talk oversamping.
Math error. Failure to select by randN. Failure to calculate margin of error. Failure to normalize against paired randR. You are not allowed to weight data by any factor. Raw data MUST be used.Lets say you live in a nation with 50 democrats and 50 republicans. You do a poll on whether or not you should ban pool noodles and you get responses from 60 democrats and 40 republicans. ALl you have to do is simply weight the republican answers higher so that you get more brought in line with your communities demographics.
No, this is what happens all the time when people conduct polls. They do not publish the raw data. They do not conform to the rules of statistical mathematics. They do not publish the questions.This is what happens all the time when people go "herr deerrr they sampled more liburals"...
The rules of statistical mathematics cares. You can't do that.yeah they sampled more of them and weighed them less who cares.
No, it's bad math.It's really not that difficult. it's very basic stuff.