Surveys on surveys on surveys

[keita]

The FAST network study released some preliminary ideas on how to make buses in the Triangle better!

…but they want comments by email, which I thought is stupid.

So I made a survey-ish thing, and y’all can be a part of it too:

https://pol.is/6khwz8rzen

“A part of it”, because you can add your own ideas for others to vote on, too. This platform was one of the tools that made Taiwan’s direct democracy initiative a success, so I figured it could help here as well.

Check back here often, and let me know what you think!! You don’t need to answer every single question on here, though you will make my life easier if you vote on more things.

I’ll run some statistics and make them into an open letter by the end of the Preliminary Results’ comment period at the end of August.

More about this survey!

@What I will (and will not) do with this data

From how I understand pol.is’ data collection methods, I would basically download a sparse matrix of binary votes on each question for each user. I can tell, for some person, who voted on which questions (e.g. how many people said they’re from Cary and would never ride a bus?).

…but I cannot tell who a voter is, beyond that. I’m not sure how voters are identified -maybe it’s by randomly-generated IDs. Voters can use Twitter or Facebook to save their responses, but I shouldn’t have access to specific handles or user IDs?- but it’s my understanding that I can’t identify specific survey voters. (e.g. I wouldn’t know who that Cary-based bus-hater is.)

Since I created a free account on pol.is to create this survey, to the best of my knowledge, I’m the only one with the ability to download the results. Pol.is is SSL-secured and was signed with a valid 256-bit encryption key, and my computers have healthcare-grade security. It’s not FBI/CIA-level, but I definitely do what I can to make sure that anything I lay my hands on are reasonably protected.

I’ll keep an eye on this poll until August 5, where I’ll download the results and run a couple of analyses.

1. Look at the distribution of votes on each question
   How many people (dis)liked a specific prompt? That’s simple enough. Grouping together similar questions could be fun as well.

2. How likely was someone to answer 2 or 3 questions in specific ways?
   Linear or multilinear regressions for binomial distributions. Duhhh.

3. Is there a pattern to how people answered specific prompts?
   High-dimension clustering of responses to specific questions (or low-dimensions of groups of questions) will be used here. This assumes enough people answered enough prompts for a meaningful analysis.

4. Contradicting answers?
   Some questions ask about the same idea, but using different wording. This will be a good way to look at the reliability and repeatability of results.

5. Demographic comparisons
   I don’t collect personal identities -but that doesn’t mean I won’t learn about what responders tell me. Specifically, I ask broadly for where people live, work, shop etc. to establish a pattern between people’s backgrounds and votes. These questions will be limited, though (e.g. I won’t ask “do you party at Legend’s often?” because I don’t care to single out LGBT+ respondents), and you can see what the questions are simply by flipping through question options.