Need to test affinity

[tbbooher]

and, um, understand the math

[Xqua]

The math is not really hard ! Lemme know if in need

[tbbooher]

yeah -- i understand the probability, but I don't understand how user's have a non-scalar F and S score. it seems they simply get a single score from the first test. the only repeated trials come from the bills, which receive multiple user votes, each with their own [F,S]

[Xqua]

There is a lot of way to approach this issue, one way is: you keep all the yay vote of the user, that gives you a distribution of F,S scores (the bills FS scores) which gives you the average and variance for the user (ideally you keep the distribution as non Gaussian but for simplicity we do)

Another way which I think is better is the OK cupid approach were you have a like and dislike score based on people who voted yay and nay on a bill. So each time you vote yay or nay you get an update on one of the two distribution ! And then you can say to the user : you are X% enemy and Y% match with that bill/user/politician

[Xqua]

As for how we implemented it during the hackathon there is almost no reupdate etc. Basically what I did was to create the function that compares two point in space

[tbbooher]

Yes, and that is a straightforward function. I'm going to implement that and a more naive strategy and deploy the site to heroku so we can start playing with it.

On Thu, Apr 2, 2015 at 1:05 AM, Xqua notifications@github.com wrote:

As for how we implemented it during the hackathon there is almost no reupdate etc. Basically what I did was to create the function that compares two point in space

— Reply to this email directly or view it on GitHub https://github.com/tbbooher/ekklesia/issues/2#issuecomment-88736611.

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Tim Booher www.theboohers.org | tim@theboohers.org 703.649.4264 (home) (703) 822-7863 (mobile)