Jump to content



Photo

Google Releases Raw Election Polling Results


  • Please log in to reply
No replies to this topic

#1 Asrokhel

Asrokhel

    Neowinian

  • 1,027 posts
  • Joined: 05-April 12
  • OS: Windows 8 Pro x64 (testing to see if I keep it or go back to Windows 7)

Posted 21 November 2012 - 12:09

Quest for the Best Visualization & Analysis

The challenge

Your quest: Find the most insightful visualization or analysis of this data. Share your results with fellow data-heads and be recognized for your skills.

All submissions will receive one coupon towards free research using Google Consumer Surveys. We'll highlight the best submissions on our G+ Page at the end to share what you have built.

Due date

12 Noon UTC, December 20th, 2012

Submit your results here

Download the data

Google Consumer Surveys Election Perception Data (CSV, 1.6 MB)
Data format described here

Our novel measurement

On election day the Google Consumer Surveys team ran a survey with two parts:

1. Who do you want to win the US Presidential Election?
2. Regardless of who you want, who do you think will actually win the Election?

The goal was to measure the progress of the election over time and how it would permeate the Internet. We collected almost 10,000 samples from unique respondents between 2am PST and 11pm PST, about one data point every 10 seconds.

Now that we've collected the dataset, we want to provide it to the community to see what interesting analyses, insights, and visualizations you can discover. This is a quest for the data scientists, engineers, statisticians, and political scientists out there — anyone with an enthusiasm for data.

You can view the survey results using our standard interface.

Example visualizations

These are visualizations we did to spark your interest in the dataset.

Interactive globe of respondents

Rendered with WebGL. We did this by taking city locations, mapping them back to latitude and longitude, and the placing them on the map. To give a sense of density, we perturbed the coordinates around each point, so many overlapping responses would look like a cluster. What you end up with is something that resembles a population density map of the United States.

Posted Image
http://consumersurve...mple/globe.html

Respondent region over time

One interesting aspect of the perception graph is the shift after the concession speech at 9pm PST. The graph seems to stabilize at a 50/50 chance for respondents who favor Romney. Why is that? This plots the respondents' general location over time. It shows how after the concession speech, respondents in the West started to be more common, presumably because people in the east had gone to bed after the speech. This highlights how time-of-day effects are important to consider.

Posted Image
http://consumersurve...on_regions.html

Perception of winner by favored candidate

Splits the respondents into four groups: Obama favored, Obama winning; Romney favored, Romney winning; Obama favored, Romney winning; Romney favored, Obama winning. At the beginning of the day, you can see each group's level of confidence in their favored candidate. As states get called you can see the perception of who will win change over time.

Posted Image
http://consumersurve...perception.html

More ideas to explore

* Animate the results on the map over time
* Annotate trends in the data with media events
* Determine when perception correlates and diverges
* Join the data with other public datasets
* Find the best predictor via regression
















http://www.google.co.../elections/data




Click here to login or here to register to remove this ad, it's free!