Survey folks: What's the most useful thing you read this year? Article, tweet thread, blog post, etc.
@kwcollins
- 481 Followers
- 65 Following
- 19 Posts
Co-founder, Survey 160. Loves survey and voter participation, causal identification. Researcher. Democrat. YIMBY. Dad. Opinion haver and measurer. National Parks enthusiast. He/Him
Of course, part of what made Twitter great was ability to interact with different communities seamlessly. I'm interested in polling, but also cooking and national parks. Not sure that Mastodon's federated structure facilitates multiple different interest-based networks very well
RCP was the third federally-created national park, established by Congress in 1890, before even Yosemite. However, while administered by the National Park System today, it is not one of the 63 official "National Parks"
Posting some of my more esoteric takes while we're on Twitter deathwatch. First up: Restore Rock Creek Park to National Park status.
And the times it's most likely to go down are the times when most people are on, aka the times when it's most interesting and valuable.
The twitter situation just kind of makes me sad. I like twitter, talking about research, making dad jokes, and discussing politics. And I have a hard time believing we won't be returning to regular Fail Whale status when twitter cuts half its staff and cuts its infrastructure
But also, polling is still the most reliable way to assess the state of races, far more reliable than, for instance, differences in early vote by partisan registration.
So look at the polls, understand them with correct caveats, and go vote regardless of what they say.
The implication of that is that given typical polling sizes, you simply can't tell who will win with certainty in close races from polls (or even lots of polls averaged together), and expecting polling to provide more certainty than it can will lead to inevitable disappointment
For state level polls, you should *at least* mentally double the margin of error to account for non-sampling sources of error. For more on this, see https://5harad.com/papers/polling-errors.pdf