Sorry to spoil the soup but calling 1000 people from California and 1000 from Wyoming and Rhode Island and ... is not good polling practice.
You have to allocate the sample size for each state based on the state population at a minimum. Dividing those population by the known D/R/I distribution is also necessary
Correct me if I’m wrong but I don’t think any of that was done.
Bottom Line: A meaningless but fun poll.
silly me, you didn't spoil my soup, but I'm sure your just an over educated individual that knows everything about everything. Seems we have a lot of folks here that fit that bill..
Have a great week end...ENJOY LIFE...
THIS AMERICAN FAMILY IN PENNSYLVANIA WILL CAST OUR VOTE FOR TRUMP~PENCE..WITH PRIDE AND HONOR..AND THIS WAS A GREAT POLL..I'M SURE IT WAS TAKEN BY CONCERNED PEOPLE WHO ARE FED UP WITH THESE MEDIA CONTROLLED POLLS.
“You have to allocate the sample size for each state based on the state population at a minimum. Dividing those population by the known D/R/I distribution is also necessary
Correct me if I’m wrong ....”
Since you ask.
Sample size and margin of error are inversely related. Bigger samples = smaller error due to chance (all things else being equal). 1,000 is a large sample. When you’re dealing with large populations (e.g. that of any state), the require sample size is the same — regardless of the population.
It’s not “necessary” to divide the sample, in advance, by R/D/I. If you have valid and reliable numbers for these subsets, then it can help to use them. But, what if you’re wrong — we are talking about making a prediction. The election is a future event, and we don’t know how it will turn out. Trying to make a good guess, is the point of this whole thing.
If you suspect a systemic bias in your sample (e.g. perhaps fewer Rs will respond to your survey), you could compare the actual distribution of the sample vs. the “predicted” one. The results can be adjusted after-the-fact, on the basis of whatever assumptions you wish to make.
Agreed. Very poor methodology, utterly meaningless, but fun.
“Sorry to spoil the soup but calling 1000 people from California and 1000 from Wyoming and Rhode Island and ... is not good polling practice.”
This is my understanding as well. If a small state is 90% Trump and a huge state is 60% Clinton, and you have polled 1000 from both, what is that supposed to prove? The big state might have 40 more electoral votes than the small state.
It also seems to me that a true poll of 1000 people from all 50 states might be useful if that info was broken down by state on the website.
Freegards