AnalyticBridge
Data Intelligence, Business Analytics
Social media to determine the next President of United States: Flawed Poll, Intrinsic Bias in Social Media
See below Attensity's poll. Using Yahoo users, and Yahoo users only, as participants in the survey will result in a significant bias. This is bad from a design of experiments point of view, and no statistician would run a poll this way. Yahoo users (and social media users in general) are more liberal than other people. For instance, many older people or very poor immigrants who have no Internet account are quite conservative. This reminds me the famous poll 80 years ago when everybody with a phone (about 1% of the population at that time) was phone-interviewed. Turned out these 1% were the wealthiest Americans and had a very different opinion from the remaining 99%.
Palo Alto: Attensity, a provider of Social Media Analytics has teamed up with Yahoo to provide Yahoo (NASDAW:YHOO) to provide millions of yahoo users to vote and guage the popular sentiment for a certain presidential candidate.
Attensity calls it The Voice of Voter Application, which measures, monitors and analyses social media buzz around the presidential candidate in real time during the course of the upcoming ABC News debate among Presidential candidates, and deliver in Real time, dynamic reports to the Yahoo! elections website at http://news.yahoo.com/elections/debate/
It will include sentiment scores for every candidate as well as performace ratings on various issues that span healthcare, defense, economy and so on.
Views: 369
Comment
-
Comment by Marcel Remon on January 21, 2012 at 9:45pm -
Another example of biased poll: Nielsen cold calling people Saturday at 8pm about a survey. Who is going to answer? Certainly not the "average" person used to detect and avoid spammers very quickly. They claim they correct for bias. But how? By scheduling 1% of they calls during business hours, and 99% outside business hours, and then by applying weights? That would still be very biased. Do they offer Internet surveys? And what about people like me who never answer a poll, because of lack of time? These people are very different from respondents.
-
Comment by Sean Flanigan on January 10, 2012 at 12:45am -
We see these "just for fun" polls on major news outlets all the time and even with the occasional caveats they seem to come across as official but they are not only biased because they are polls among the viewers (biased subset), but they are also the opinions among those who bothered to respond to the polls (another biased subset). What is very confusing is when firms trying to position themselves as analytics experts use such poor methodology without the caveats. One of my favorite polling companies is Harris. They use top notch methods and are able to articulate that even the top notch methods are not perfect. Great post!
Top Content
Follow us
Big Data Jobs
© 2013 AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC
You need to be a member of AnalyticBridge to add comments!
Join AnalyticBridge