Yellowstone From United States of America, joined Aug 2006, 3071 posts, RR: 7 Posted (3 years 4 months 2 days 6 hours ago) and read 1601 times:
Nate Silver over at Fivethirtyeight (now part of nytimes.com) has done his usual post-election investigation into how accurate the pollsters were at forecasting the election, and it turns out that Rasmussen Reports didn't do a very good job:
- Their polls were an average of 5.8 points off from the actual margin of victory.
- One poll (of the Hawaii Senate race) missed the mark by a full 40 percentage points, the largest error in a political poll since 1998.
- Their polls had a 3.9 point Republican bias, relative to the actual results.
So yeah, next time you see a Rasmussen poll, take it with a few extra grains of salt!
LTU932 From Germany, joined Jan 2006, 13864 posts, RR: 51
Reply 1, posted (3 years 4 months 1 day 16 hours ago) and read 1512 times:
In Germany we have a saying: "Vertraue niemals einer Statistik, die du selbst nicht fälschen kannst" (Don't trust a statistic you can't manipulate yourself).
The point is: Take ALL polls and ALL statistics with a few barrels of salt. One example are the unemployment numbers in Germany, where officially unemployment dropped to under 3 Mio, yet the statistic omits certain other people registered as unemployed (e.g. unemployed people over 50, people who have to get by with jobs that pay only 1 Euro an hour, etc.), which means that the actual number (the one the government rejects as outright wrong) is around 4.5 Mio.
dxing From , joined Dec 1969, posts, RR:
Reply 3, posted (3 years 4 months 1 day 15 hours ago) and read 1490 times:
The story itself seems a little biased. The last line says that CNN while almost as wrong as Ramussen only released 14 of 17 races versus Ramussens 105. So Ramussen put out an additional 88 polls and someone is suprised that their error factor was higher? The other pollsters cited also put out a smaller number of polls. It is like saying that an airline that only flies to 10 cities has a better on time average than an airline that flies to 100 cities. When the amount of polls taken is equal how do they fair? How about a match up of Quinninpac and Ramussen head to head on the same polls, what was the error factor then? Not even mentioned. Seems like sour grapes to me.
Rasmussen polls were showing Sharron Angle leading by as much as 8 points over Harry Reid. Same for the Patty Murray / Dino Rossi matchup in Washington and Bennett vs. Buck in Colorado.
The Republicans lost in all of those races.
Quoting MD-90 (Reply 2): Polling by calling landline phones is increasingly irrelevant in this cellular age.