BluePhantom
Educator (of liberals)
The purpose of this thread will be to provide an analysis of polling through the election in November. I have a spreadsheet that I wrote that crunches the numbers very effectively and has shown some very solid accuracy in past presidential, gubernatorial, and congressional elections. While it has the ability to get pretty indepth with its calculations I am going to avoid that for now because a) the election is too far away to be focusing so closely on fractions of a percentage, b) some of the calculations would require far too much space writing out equations, and c) all that is more "my personal method" while the intent is to take a more general view here.
As much as possible I will attempt to be non-partisan and ask you to do the same. It's my hope that we can engage in analysis from an objective point of view in order to reach conclusions about what the data actually suggests instead of what we are hoping it says.
The procedure I will use on this thread is as follows:
1) All polls will be included whether I, or anyone else, finds that polls to be reliable or not.
2) Polls will be averaged out to provide baseline data into three categories: a) professional polling organizations, b) media polls, and c) all polls combined. I will post polling averages weekly in these three categories as well as a two week average.
3) Polls that were not completed before April 10th will be ignored. This is because that is the day that Santorum dropped out of the race and the entire landscaped changed after that. The first weekly cutoff date will therefore be April 17th, the second will be April 24th and so on. Polls will be categorized according to the date they were completed.
4) Daily tracking polls that use a rolling average (Rasmussen and Gallup, for example) will be calculated according to their current data on the final day of the weekly period to avoid over-representation within a given week. However, for the two week average and for purposes of identifying trends they will be considered separate polls from one week to the next. This will avoid redundancy since Rasmussen uses a three day rolling average and Gallup a five day rolling average...in other words enough time will have passed in their averages to consider them separate polls from one week to the next as previous data will have dropped off their averages.
5) Polls over two weeks old will be completely ignored as the data is too old to consider except where trends can be identified. Personally I prefer 10 days but we will go with 14 for ease and consistency on this thread.
6) And I want to stress this again.....I encourage as much objective analysis and debate about interpretation as possible. However, those who wish to cry and moan about this poll or that poll and turn this into a political debate are strongly encouraged to opine elsewhere. I encourage everyone to read the OP in Poll Reading 101 for some basics in polling analysis.
The first set of data on the week ending April 17th will be posted shortly.
As much as possible I will attempt to be non-partisan and ask you to do the same. It's my hope that we can engage in analysis from an objective point of view in order to reach conclusions about what the data actually suggests instead of what we are hoping it says.
The procedure I will use on this thread is as follows:
1) All polls will be included whether I, or anyone else, finds that polls to be reliable or not.
2) Polls will be averaged out to provide baseline data into three categories: a) professional polling organizations, b) media polls, and c) all polls combined. I will post polling averages weekly in these three categories as well as a two week average.
3) Polls that were not completed before April 10th will be ignored. This is because that is the day that Santorum dropped out of the race and the entire landscaped changed after that. The first weekly cutoff date will therefore be April 17th, the second will be April 24th and so on. Polls will be categorized according to the date they were completed.
4) Daily tracking polls that use a rolling average (Rasmussen and Gallup, for example) will be calculated according to their current data on the final day of the weekly period to avoid over-representation within a given week. However, for the two week average and for purposes of identifying trends they will be considered separate polls from one week to the next. This will avoid redundancy since Rasmussen uses a three day rolling average and Gallup a five day rolling average...in other words enough time will have passed in their averages to consider them separate polls from one week to the next as previous data will have dropped off their averages.
5) Polls over two weeks old will be completely ignored as the data is too old to consider except where trends can be identified. Personally I prefer 10 days but we will go with 14 for ease and consistency on this thread.
6) And I want to stress this again.....I encourage as much objective analysis and debate about interpretation as possible. However, those who wish to cry and moan about this poll or that poll and turn this into a political debate are strongly encouraged to opine elsewhere. I encourage everyone to read the OP in Poll Reading 101 for some basics in polling analysis.
The first set of data on the week ending April 17th will be posted shortly.
