Posts by David Hood
Last ←Newer Page 1 2 3 4 5 Older→ First
-
Hard News: Gower Speaks, in reply to Peter Green,
I was actually adding some code for loading gtools while you were writing that :)
-
And here it goes
https://github.com/thoughtfulbloke/OneVoteTwoVoteRedVoteBlueVote -
Hard News: Gower Speaks, in reply to JonathanM,
Jonathan, Great (and a classic demonstration that posting something bad on the internet is a great way to get something better).
Since I was going to add it to github, but you have done a lot of work on making it actually work, I was going to mark the project MIT licence (a basic attribution for using the code in other things) or Public Domain, do you have an opinion?
-
And with the Roy Morgan anything but NZ First being a decider is pretty unlikely (baring a Conservative seat, and even if the Conservative's have one NZ First deciding is still more likely than anything else by a bit).
Minor correction to the above post- with the Colmar results, in my very bad simulator, NZ First is about as likely to be the decider as National is to be clearly ahead.
-
Here is a very, very rough election result simulator (written in R) just using random binomial draws based on the support in a poll.
https://www.dropbox.com/s/ylbsp9ggm1y9lls/oneVoteTwoVoteRedVoteBlueVote.R
I'll put up a proper copy on github tomorrow, but my computer this evening isn't really set up for that.
With the Colmar poll, the single most likely outcome (49% ish ish ish) was that it was up to NZ First provided the Conservatives were not gifted a seat. If the Conservatives have a seat the single most likely result is a clear National victory (54% ish ish ish).
But this is so speculative it falls into "for the purposes of entetainment".
Also, while I know data, I am sooo not a statistician. -
I think, because of the 5% line, if drawing virtual Parliament results, you kind of need two, side by side, with some visual indication of how likely each style of results, because the flow on effects of NZ First's result are (I think, looking at the numbers) going to have a more significant effect on government formation than anything else. Once NZ First's effects are dealt with, the other perturbations due to error ranges are pretty minor.
-
I was just reading Tim Harford's high accessible piece on "big data" in FT Magazine, but while it is in theory about big data it includes good discussions of polling biases and other media-stats relevant issues:
http://www.ft.com/intl/cms/s/2/21a6e7d8-b479-11e3-a09a-00144feabdc0.html#axzz2xRsgyMWe -
State of the art in automatic sentiment analysis is (what I would describe) as moving from "deeply flawed" to "marginally useful" these days. What you would need are a bunch of transcripts (so this favors written media as a source of analysis).
Another approach would be to employ a random collection of survey takers for a few dollars on Amazon's Mechanical Turk (to avoid observer bias in classifying) asking the question "rate how positiviely the media treated the subject in this clip". But that assumes anyone is interested enough to fund a survey.
I tend to think the crowd-sourced corpus tend to work better for assembling facts than assessing opinion, as they are two susceptible to where the crowd is coming from.
-
Of my (almost 16) year old daughter, every teacher since she started intermediate has remarked that her year (not just her) is vastly more politically aware than other years, the year above somewhat but her year in particular. So I would suggest whatever was happening in formative education a decade ago was good stuff.
-
Related to zooming in on little things, I thought this looked like it might be interesting when there was news about it a few weeks ago
http://boingboing.net/2014/03/10/bioengineer-builds-50-cent-pap.html