5 ms·
It's easy to think that but it's not true. Regardless of population size, providing a sample is truly random, a sample size of 300 is enough to measure with an
by shalmanese 13y ago
It's easy to think that but it's not true. Regardless of population size, providing a sample is truly random, a sample size of 300 is enough to measure with an accuracy of + or - 3%. Increasing sample size much beyond this does not meaningfully increase accuracy.
An analogy I've found useful thinking about this is, if you have a pot of soup, tasting a teaspoon is enough to figure out whether the soup is oversalted or not, regardless of the size of the soup pot, providing the soup is well mixed.
- Link- 13y agoAssuming the set is homogeneous! Which in this case, is not. A sample extracted from a more or less financially stable region of the country will have drastically different output than a sample extracted from a poorer area. Your assumption about 'the soup well mixed' is less probable to be true, given the lack of any evidence supporting the claim that this study is either meaningful or properly executed.
- shalmanese 13y agoRight, if there's a flaw in the study, it's in the representativeness, not the sample size.