Graphs of probability distributions

[endolith]

Currently you show a bunch of symbols of different sizes and colors, with each representing an average of many simulations, correct? But it's pretty cluttered and hard to see general trends.

Could you just plot the voter happiness directly as a probability distribution for each system? (Does it make sense to refer to the VSE of a single simulation?)

Kind of like http://electology.org/sites/default/files/comparing_voting_methods_simplicity_group_satisfaction.png

but more like this:

or maybe beanplot style:

[endolith]

Also, what changed between http://rpubs.com/Jameson-Quinn/vse3 and http://electology.github.io/vse-sim/vse.html ? You say V321 is better than SRV, but in the older plots it doesn't look like it.

[jamesonquinn]

What changed is that in the older graph, strategic voters were using a different strategy; one that was more likely to backfire and less likely to work. Backfiring strategies made VSE worse for everyone.

2017-02-14 20:47 GMT-05:00 endolith notifications@github.com:

Also, what changed between http://rpubs.com/Jameson-Quinn/vse3 and http://electology.github.io/vse-sim/vse.html ? You say V321 is better than SRV, but in the older plots it doesn't look like it.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/electology/vse-sim/issues/17#issuecomment-279894029, or mute the thread https://github.com/notifications/unsubscribe-auth/AAWu9GrWMZRkIO8CQTLyq06MfMEWFgqbks5rcllHgaJpZM4L6bkU .

[jamesonquinn]

Yes, I could plot distributions like that. (I don't immediately know how, but I'm sure I could manage it if I needed to.) But there would be separate distributions for each strategic assumption, and that would make the graph either less informative or just as cluttered.

2017-02-08 1:26 GMT-05:00 endolith notifications@github.com:

Currently you show a bunch of symbols of different sizes and colors, with each representing an average of many simulations, correct? But it's pretty cluttered and hard to see general trends.

Could you just plot the voter happiness directly as a probability distribution for each system? (Does it make sense to refer to the VSE of a single simulation?)

Kind of like http://electology.org/sites/default/files/comparing_ voting_methods_simplicity_group_satisfaction.png

but more like this:

[image: probdist] https://cloud.githubusercontent.com/assets/58611/22725779/7536688c-ed9d-11e6-935b-d293b007590a.png

or maybe beanplot style:

[image: figure-4-region-state-interaction-and-party-politics-notes-bean-plot-with-density-trace] https://cloud.githubusercontent.com/assets/58611/22725785/7d29fd7e-ed9d-11e6-9d23-f97151ba9a83.png

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/electology/vse-sim/issues/17, or mute the thread https://github.com/notifications/unsubscribe-auth/AAWu9Br-y2oR0BZ0oikX-wxMKTGvJ4Fjks5raWAtgaJpZM4L6bkU .

[endolith]

I'm saying to combine all the strategy mixtures and scenarios into a single distribution for each voting method, without lumping groups of elections together. I assume that within each scenario and strategy grouping, there are multiple simulations with a spread of results, right?

Do you generate the elections first and then separate out the ones that fit different scenarios, or do you intentionally generate different scenarios directly?

[jamesonquinn]

The scenarios are artificially separated. But the strategy mixtures are set a priori. If I wanted one number for a given method, I would do it by setting strategy percentage based on strategy effectiveness. This is hard to do, though, and any actual implementation would involve even several times more arbitrariness in parameter-setting than what I'm doing so far.

2017-02-15 20:06 GMT-05:00 endolith notifications@github.com:

I'm saying to combine all the strategy mixtures and scenarios into a single distribution for each voting method, without lumping groups of elections together. I assume that within each scenario and strategy grouping, there are multiple simulations with a spread of results, right?

Do you generate the elections first and then separate out the ones that fit different scenarios, or do you intentionally generate different scenarios directly?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/electology/vse-sim/issues/17#issuecomment-280196235, or mute the thread https://github.com/notifications/unsubscribe-auth/AAWu9D92qfl1COhZm_reqo7YT94N7yfhks5rc6EvgaJpZM4L6bkU .

[endolith]

(This was my attempt, by the way:

)

[dylanhs]

Can you add labels to the axes on that graph? Is that Bayesian Regret on the abscissa?

[endolith]

@dylanhs My graph? The X axis is the distance from the winner to the center of the population. I think it might be more clear if it were symmetrical and X axis were just position of winner and candidates were uniformly distributed