I compared how two different sampling methods (random walk and forest fire) "perform" on two different data sets (a road network from Texas, and a social graph from Facebook), by comparing different network measures on the samples.
There is much room for improvement; in fact, the differences could just be due to my implementing some parts of the algorithms wrong.
All code and text are unlicensed
The data sets are taken from Konect, and thus subject to the Creative Commons Attribution-ShareAlike 2.0 Germany License.