Updates and thoughts.
Others can discuss elsewhere the broader implications of Brexit. I want to make a few brief points about the Brexit polling.
This is the first in a series of posts analyzing tweets about the election.
We use a multilevel regression and post-stratification (MRP) approach to obtain estimates of voting intent.
Petit Poll uses biased samples. Traditionally, this would make our results meaningless, but we are able to make reasonable conclusions because we tie our samples to representative data.
Unless you live in Canberra you are unlikely to be exposed to our Facebook/Twitter ads. So I thought I'd take a moment to show you what they look like.
Watching Mad Men is the extent of my marketing experience. But n (the sample size) is an unavoidable aspect of confidence intervals. So I need more respondents than just my family and friends. This means advertising.
Maintaining the trust of people who participate in Petit Poll is a critical aspect of operating Petit Poll. I would prefer to shut down Petit Poll than undermine that trust. Please point out any mistakes that I am making or areas that I could improve.
Sorry for the silence for the past few days.
We just ran the model on the tiny sample that we were able to gather from friends and family. There are a huge number of issues that we still need to solve, but the process seems to work.
We started our Facebook ad process last night.
Beta-email is a group of friends and family who signed up for our polls.
Petit Poll's Twitter account is live.
I got onto the plane (enplaned?) in Sydney an hour or so before Turnbull went to see the GG, which meant that when I landed (and deplaned?) in San Francisco the election was well underway.
In a few hours Malcolm Turnbull is expected to announce an election. This will be Petit Poll's first outing and our initial surveys are ready to go.
Petit Poll's version 0.1 first test delivered the hoped for result.