Now I am laying out the logistics on what I am going to do this weekend. I will begin running states today for the last time. Here is my plan a rough order of calls. Utah will be considered a red state, because I am no longer concerned with the possibility of a Trump loss. The swing states will be divided into two groups depending whether the race is close or not. Washington DC data will be manually collected from the one pollster doing semiscientific samples (SurveyMonkey), and a csv file will be created. Washington D.C. will be treated like a blue state. Pollster does not include the Google Mobile Surveys, and I am not going to add those because I don't like how they are run (they don't get a good fair sample).
The order is as follows
1. Red States (Data pulled Friday at around 3 pm CST)- Alabama, Alaska, Arkansas, Georgia, Idaho, Indiana,Kansas, Kentucky, Louisiana, Mississippi, Missouri, Montana, Nebraska, North Dakota, Oklahoma, South Carolina, South Dakota, Tennessee, Texas, Utah, West Virginia, Wyoming
2. Blue States (Data pulled either Friday at around 6 pm or 9 am Saturday morning) -
California, Connecticut, Delaware, Hawaii, Illinois, Maine, Maryland, Massachusetts, Minnesota, New Jersey, New Mexico, New York, Oregon, Rhode Island, Vermont, Washington , Washington D.C.
3. Non-Close Swing States (Data pulled Saturday after 10 am) - Colorado , Michigan, New Hampshire, Pennsylvania , Virginia, Wisconsin
4. Close Swing States (Data pulled Saturday after 3pm), in order of least close to closest race - Arizona, Iowa, Florida, North Carolina, Nevada, Ohio
Note : Ohio and Nevada are close and they could flip from my Saturday prediction (which will probably be a Clinton win in both states) if more polls showing a Trump lead appear before election day. I might create an update on Sunday or Monday if the prediction changes. These are probably the only states that could flip after my Saturday prediction. Ohio is more likely to change under my model than Nevada.
The date and time of when the data is pulled will be included in the analysis title for each state. There will be 4 different txt files and each one will only have the states in each group. The summary post will include a map of all 50 states plus Washington D.C.
The final prediction will involve 5 posts, one from each category, and a summary post. All posts will be on Saturday even if the analysis was completed on Friday. My method of winning the real winner with overlapping intervals will to first reduce the interval size to 90% , and if that is still inconclusive hypothesis testing will be done. If I feel that my model is wrong in one or more states I may provide another map with less statistically based predictions. These predictions not based on my model will not be included in determining the success of my model. If new polling comes available in any of the 6 close states on Sunday or Monday, a new analysis may be run and only if the winner is different will the new analysis be posted. My main measure of success is how my Saturday model compares to election day. While I expect that my model should provide around 95% success (47 or 48 states), it may not work or it could work very well and predict all the states. Variation is a part of statistics. To help compare I will look at the predictions made by Princeton Election Consortium, FiveThirtyEight, and 270 to Win at the time my model is complete.