Statistics is not intuitive and weird. If the population of people asked were randomly chosen in a representative manner, then 1 thousand people is definitely enough for an accurate reading for a country of 40 million people
You only need 1068 people to achieve a 95% confidence interval (95% of answers are expected to be the same if you redid the survey, also the industry standard to achieve) and 3% margins of error for a population of 40 million
Play around with the calculator a little bit and you’ll find something interesting. 2400 people are needed for a 95% interval of the population of USA (300 million). However, if you increase that population size from 300 million into even dozens of billions, the sample size required to achieve 95% confidence increases slower and slower to the point where it’s almost useless to increase your sample size past a few thousand
That is correct, but it’s always important to make sure the “randomly chosen” bit is close to true, which is usually true (and verifiable if the methodology is accessible) in electoral polls and such, but I would hazard a guess that this is not as obviously the case in a country having a civil war, specially with an agency that goes out of its way to call one of the positions as “Russian propaganda.” Their overview on methodology doesn’t mention in which regions the polling was made and their (google doc) source is in Ukrainian/Russian and I can’t read that.
But yeah, in general people are often too quick to dismiss surveys because of sample sizes that are unintuitively small.
Yes, but at the same time that’s mathematical statistics. You still have to contend with countless types of biases that skew results in plenty of ways.
Where did they get these people from? The Donbas or Kiev? I’m sure they would have different answers. What ages are the participants? What are their previous political affiliations? Why did they remain in the country when millions of other fled?
Not to mention the biases of the information taker, as those can be nefariously purposeful, or also accidental.
Getting statistics to be as clean as they are in the math will always be next to impossible.
Statistics is not intuitive and weird. If the population of people asked were randomly chosen in a representative manner, then 1 thousand people is definitely enough for an accurate reading for a country of 40 million people
You only need 1068 people to achieve a 95% confidence interval (95% of answers are expected to be the same if you redid the survey, also the industry standard to achieve) and 3% margins of error for a population of 40 million
Play around with the calculator a little bit and you’ll find something interesting. 2400 people are needed for a 95% interval of the population of USA (300 million). However, if you increase that population size from 300 million into even dozens of billions, the sample size required to achieve 95% confidence increases slower and slower to the point where it’s almost useless to increase your sample size past a few thousand
https://www.checkmarket.com/sample-size-calculator/
That is correct, but it’s always important to make sure the “randomly chosen” bit is close to true, which is usually true (and verifiable if the methodology is accessible) in electoral polls and such, but I would hazard a guess that this is not as obviously the case in a country having a civil war, specially with an agency that goes out of its way to call one of the positions as “Russian propaganda.” Their overview on methodology doesn’t mention in which regions the polling was made and their (google doc) source is in Ukrainian/Russian and I can’t read that.
But yeah, in general people are often too quick to dismiss surveys because of sample sizes that are unintuitively small.
Yes, but at the same time that’s mathematical statistics. You still have to contend with countless types of biases that skew results in plenty of ways.
Where did they get these people from? The Donbas or Kiev? I’m sure they would have different answers. What ages are the participants? What are their previous political affiliations? Why did they remain in the country when millions of other fled?
Not to mention the biases of the information taker, as those can be nefariously purposeful, or also accidental.
Getting statistics to be as clean as they are in the math will always be next to impossible.