The 8000 respondents, from 16 different countries, answered questions on what they thought about Hamas, their opinion on Iran and Western countries, and what actions were needed by Arab states.
Whilst it’s quite amazing just how few people you need in your sample to get a +/- 3% error with a 95% confidence for the opinion of hundreds of millions, that narrow error margin with just a few thousand samples only works if the sample is representative of the population in general, which is unclear.
Plenty of cases of polls out there that are complete total bollocks because they were taken by calling by phone people in relativelly poor countries were only those in the middle class and above have a phone, and at times when most people were out working, so they ended up sampling an atypical subset of people rather than one representative of the whole society.
The way the question is posed also influences the results, sometimes quite subtly (the mere order or words or order of questions in multiple-question poll can sway the results).
So whilst what you said makes sense in response to the previous commenter’s point, there’s a lot more to good polling than merelly the number of samples necessary for a certain error margin in the 95% confidence interval when your sampling is random in a uniform distribution (emphasys on the later, as actually making sure the sampling is indeed like that in practice is often difficult and/or costly)
You definitely haven’t taken stats.
Just like the polls done by those who “know stats” who predicted a clinton win in 2016.
Whilst it’s quite amazing just how few people you need in your sample to get a +/- 3% error with a 95% confidence for the opinion of hundreds of millions, that narrow error margin with just a few thousand samples only works if the sample is representative of the population in general, which is unclear.
Plenty of cases of polls out there that are complete total bollocks because they were taken by calling by phone people in relativelly poor countries were only those in the middle class and above have a phone, and at times when most people were out working, so they ended up sampling an atypical subset of people rather than one representative of the whole society.
The way the question is posed also influences the results, sometimes quite subtly (the mere order or words or order of questions in multiple-question poll can sway the results).
So whilst what you said makes sense in response to the previous commenter’s point, there’s a lot more to good polling than merelly the number of samples necessary for a certain error margin in the 95% confidence interval when your sampling is random in a uniform distribution (emphasys on the later, as actually making sure the sampling is indeed like that in practice is often difficult and/or costly)