1. Are you totally obsessed w the #GASenateRunOffs? Check out this amazing tool from @OldBullTV & @OpenModelProj that let's YOU take a deep, interactive dive into their survey data to see the impact of diff weights & turnout models so you can understand how polling data is VERY
2., dependent on the decisions made by survey researchers. Everything from the Qs included in the polls (AND how those Qs get asked AND how respondents are allowed to respond) to the "under the hood" stuff like the weights used for things like age, race, gender, & education.
3. Most people do not understand that these decisions are imposed on the data and that the results then, come out a certain way. This means that one researcher imposing one set of decisions, based on one set of assumptions could analyze the same data as another researcher and get
4. different results!! This is sort of what happened in 2017 with my polling in the VA Gov race compared to nate Cohn's poll. I anticipated an entirely new VA electorate- one that would be more female, non-white, younger, and more educated than any that had existed in the past &
5. relied on the traditional method of looking at the previous comparable cycle to estimate what the electorate would look like, which is all well and good unless the electorate is about to change profoundly. This is why our relationship got off to a rough start, btw. When I tell
6. people that the only polling data that I really "trust" is my own bc then I'm privy to the full process of how it was weighted and modeled I don't mean it the way people think I do. I mean that least I know precisely what the survey researcher biases are! So, spend some time
7. with this fab open source tool which will be a GREAT teaching tool for helping people understand this fairly complex, but critical topic. Its a bit complex, so I've tried to link directly to the most relevant page openmodelproject.org/inhouse/
• • •
Missing some Tweet in this thread? You can try to
force a refresh
