sampling - this is just who was called & responded (not the same thing. e.g. to get 1000 people, you might call 4000). What you care about is the total #. That's your sample size . As you might imagine - you get a better guess with a bigger sample size
-
Prikaži ovu nit
-
weighting - this is just how you add up responses in you sample. You could just add it all as is but unless your sample looks just like the overall population - that won't be meaningful.
1 reply 0 proslijeđenih tweetova 3 korisnika označavaju da im se sviđaPrikaži ovu nit -
that's why polls often as a bunch of questions about your age, race, gender, etc. So on the backend, responses can be adjusted to closely mirror the general population (rather than who just responded).
1 reply 0 proslijeđenih tweetova 3 korisnika označavaju da im se sviđaPrikaži ovu nit -
So rather than interpreting a poll as the opinion of the 1000 people who responded, you can interpret is as the best guess on how everyone will act. E.g. - the % tells you what the like percent of the population that will vote for a candidate.
1 reply 0 proslijeđenih tweetova 3 korisnika označavaju da im se sviđaPrikaži ovu nit -
Getting the sample big enough to appropriately adjust it so it is close to the population is what makes that guess more likely to be accurate. So if you only had 10 people and 1 of those people had to represent all young women in metropolitan areas in the US, it'd be a bad guess
1 reply 0 proslijeđenih tweetova 4 korisnika označavaju da im se sviđaPrikaži ovu nit -
So how do you know how good the guess is? MARGIN OF ERROR. The margin of error tells you the range that is statistically meaningful given all the sampling and adjustments. So if the number reported is 50% and the margin of error is +/- 50% - it's not meaningful.
1 reply 0 proslijeđenih tweetova 3 korisnika označavaju da im se sviđaPrikaži ovu nit -
why do you care? suppose you want to compare candidates see who's doing better (or who is electable?) you need to know if the predicted % of people voting for the 2 candidates are different.
1 reply 0 proslijeđenih tweetova 3 korisnika označavaju da im se sviđaPrikaži ovu nit -
If the numbers are 21% & 25% w margin of error is +/- 2% they are. If the margin of error is +/- 5% they are STATISTICALLY THE SAME. One more time - if the numbers are within margin of error, even if the percentages reported look different, that difference is not meaningful
1 proslijeđeni tweet 6 korisnika označava da im se sviđaPrikaži ovu nit -
Often for primary polls with small samples, selection bias in who is responding, and a large number of potential candidates, margins of error are large and point spreads that seem big (3-5 points) are not meaningful bc the margins of error are bigger
1 reply 0 proslijeđenih tweetova 4 korisnika označavaju da im se sviđaPrikaži ovu nit -
(for detailed analysis on this, see this comprehensive 538 article https://fivethirtyeight.com/features/the-polls-are-all-right/ …)
1 reply 1 proslijeđeni tweet 5 korisnika označava da im se sviđaPrikaži ovu nit
TL;DR - there will be A LOT of polls released which purport to say who is likely to win. Before you talk about results-check the margin of error. If it's bigger than the % difference, repeat to yourself (and if you don't mind anyone around you) THERE IS NO STATISTICAL DIFFERENCE
-
-
Odgovor korisniku/ci @RadhaKIyengar
& at least W.R.T. Iowa, isn't this all compounded by the fact that there's not a hell of a lot of mapping between polls & caucus results?
0 replies 0 proslijeđenih tweetova 0 korisnika označava da im se sviđaHvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.