3 ms·

> You're depending on 1018 people who actually may not be statistically relevant, since America is a big, big country, to drive your thinking on that country.

by IanCal 11y ago

> You're depending on 1018 people who actually may not be statistically relevant, since America is a big, big country, to drive your thinking on that country. It's a problem if they're badly sampled, but the sample size you need is doesn't really change as you increase your population, as long as your population is a lot bigger than your sample size. If 1000 people would be OK for a population of 1M, then it should be just as good for a population of 100M. I don't think this really invalidates your point, it's just a misconception I see a lot. http://www.surveysystem.com/sscalc.htm http://www.surveysystem.com/sscalc.htm

russnewcomer 11y ago
You are correct, and what I should have said was something closer to "You're depending on the answers from 1018 people to be statistically relevant, and they may not be because by the end of the survery people just hurriedly say whatever to get the survey over with, and I'd bet that 50% of the answers should probably be discounted, so you're depending on what may not be a statistically relevant survery, since America is a big, big country, to drive your thinking on that country."
jndsn402 11y ago
I don't disagree with the math, but wouldn't it be the case that it's easier to get a biased sample from 1000/100M people than 1000/1M people? For instance, maybe your sampling ignored a particular region of the country or socioeconomic strata etc.
- chimeracoder 11y ago
  > I don't disagree with the math, but wouldn't it be the case that it's easier to get a biased sample from 1000/100M people than 1000/1M people? It's not the population size that matters; it's the variance in the underlying population. Mathematically, the sample size needed to construct an n% confidence interval for a normal distribution is independent of the population size (the first derivative is zero).
humanrebar 11y ago
> then it should be just as good for a population of 100M Assuming an normal (or similar) distribution. It seems plausible to me that we're in some sort of multi-modal period in our history. It may not be sophisticated enough to just take a mean or group opinions into five buckets on a one-dimensional graph.

by IanCal 11y ago

russnewcomer 11y ago

You are correct, and what I should have said was something closer to "You're depending on the answers from 1018 people to be statistically relevant, and they may not be because by the end of the survery people just hurriedly say whatever to get the survey over with, and I'd bet that 50% of the answers should probably be discounted, so you're depending on what may not be a statistically relevant survery, since America is a big, big country, to drive your thinking on that country."

jndsn402 11y ago

I don't disagree with the math, but wouldn't it be the case that it's easier to get a biased sample from 1000/100M people than 1000/1M people? For instance, maybe your sampling ignored a particular region of the country or socioeconomic strata etc.

chimeracoder 11y ago
> I don't disagree with the math, but wouldn't it be the case that it's easier to get a biased sample from 1000/100M people than 1000/1M people? It's not the population size that matters; it's the variance in the underlying population. Mathematically, the sample size needed to construct an n% confidence interval for a normal distribution is independent of the population size (the first derivative is zero).

humanrebar 11y ago

> then it should be just as good for a population of 100M Assuming an normal (or similar) distribution. It seems plausible to me that we're in some sort of multi-modal period in our history. It may not be sophisticated enough to just take a mean or group opinions into five buckets on a one-dimensional graph.