Reply to thread

Message: <blockquote data-quote="Cadence" data-source="post: 9107114" data-attributes="member: 6701124">Do you have a good link explaining this?   I'm wondering what assumptions on the opt-in mechanism are needed and/or which parameter [pct for which population] are they within 5% for. --  I'm kind of thinking of the Literary Digest poll of 1936.For most of the margin of error calculations for random samples, doesn't population size [Edit: or ratio of sample to population size] only come in when accounting for the error between the actual finite (like hypergeomtric) calculation and the commonly used infinite one (like binomial) - unless maybe something about the sampling mechanism depends on the the population size?  At the other extreme, you could end up a guaranteed 10% off even if you had 90% of the population if responding is most negatively associated with one of the views, right?And even then (except for a random sample) is there a real way to show from the one sample ho representative it is in terms of the variables they actually care about?   If the wizard lovers from demographic combination X are motivated to reply and the others from X aren't, then it doesn't matter how close they got to matchng demographic X overall, right?  -- Thinking of the 1948 quota sampling picking Dewey.</blockquote>