So, I’ve got a lot of blog posts that I meant to publish last week, but I never got around to it.  Here is a graph I made using the the auto-complete terms from Google, Yahoo, and Bing for republican presidential candidates.  I looked at the five top auto-completes from each site and scored each word 5 points if it was the first auto-complete, 4 points for second auto-complete, etc.  I did a search for each candidate twice on each site.  First using just the candidates name and a space, then the candidates name followed by the word “is” and then a space. (For example, “Mitt Romney ” and “Mitt Romney is “).   I then weighted the search engines based on their market share (about 75%, 15%, and 10% respectively).  This gives me a data set with 8 observations (8 candidates) and several dozen variables (one variable for each word).  I then used mutli-dimensional scaling to reduce the distances between the vectors down to, in this case, three dimensions.  The size of each circle is proportional to the polling percentage from RealClearPolitics on August 29, 2011 (the same day as the auto-completes were done.)  The word appearing in or next to each circle, is the word with the highest score for each candidate.

Also, one of Michele Bachmann’s auto-complete terms on Google is “slavery”.  I couldn’t imagine what she had done to warrant this as an auto-complete term, but then I found this article by Andrew Gelman (of the blog Statistical Modeling, Causal Inference, and Social Science).  Yikes.

Cheers.