Quote:
Originally Posted by Trolly McTrollson
The essential difference here is that goofy is representing the data with a confidence interval for each candidate instead of a global margin of error for the entire poll. It's much easier for a layman to see that there's no overlap between the 05% intervals for 10th place and 11th place.
Yes this is obvious but i did the calculations for the original poll data she presented in her show with corrected errors and they were very close and i estimated over 75% chance one of the 11th,12th,13th or even 9th to be the real 10th. In this chart though this is more like less than 1% though. This is why i asked if this is a new set of polls.
In the original poll dataset she made 2 bad errors (she used standard maximum error defined at 50% probability as 3.5-4% for all and she failed to update it in fact down to 1.8% after adding 5 polls that each was around 3-4% - Both those errors are terrible actually for a person that deals with polls and some simple statistical analysis often as part of her job and as someone that ought to have prepared - and her team- for her argument with more scrutiny than a random live comment). However, moreover the errors, in the original poll data her argument still had merit because they were like 2.6% vs 2.2%,1.8% 1.4% each with close to 0.5-0.6% errors. Here the 10th is more like 4%, so much higher than before it seems to make the top 10 a lot more clear and secure. I think goofball is updating ie uses what the polls she used look today (or other recent similar polls) not then.
This is the original data set she used
http://www.msnbc.com/rachel-maddow-s...r-492254275819
Here are some polls in recent times;
http://www.realclearpolitics.com/epo...tion-3823.html
Last edited by masque de Z; 08-06-2015 at 08:42 AM.