While a graph from BlogPulse allows one to see the attention around certain keywords - as in the example below for 'obama' and 'clinton', it doesn't allow one to compare how these terms are doing relative to each other.
The graph below takes the percentage values from the above and plots them as ratios (Obama as a ratio of Clinton buzz and Clinton as a ratio of Obama buzz).
What we can see here is that while certain parties may like to characterize the battle for nomination as a some you win, some you lose affair, the battle for attention appears to have a clear trend.
Note: producing this type of chart is pretty straight forward. The html page that presents the BlogPulse trend results includes client side image map data with the percentages included as labels. One can transform this in to appropriate tab separated data for a spreadsheet and then compute the ratios.
Update: here is the logarithmic version of the chart as suggested by Moritz Stefaner.
Hi, interesting analysis. And what a great tip with the BlogPulse image map!
One thing though: I was a bit puzzled for a second why the second graph is _almost_ symmetrical, and then realized that it nicely demonstrates an old trap: Plotting ratios on a linear scale underrepresents ratios < 1! If, e.g. Obama=2*Clinton for a given time point, you get values of 2 and 0.5. Now clearly, 2 is much further from the "baseline" 1 than the .5, which results in a flattened impression on the smaller values. So as a tip for improvement, I would propose a logarithmic scale - or just plotting one of the two lines since they express the same thing anyways?
So, feel tuftified :)
Posted by: Moritz Stefaner | February 17, 2008 at 12:17 PM