XKCD makes fun of the fact that a 95% certain test is to 5% uncertain, and so, if one just often enough tests, a false positive will at some point result. Ha ha ha, good message, but I have to poop on the party, sorry – party pooper – don’t know me yet? I mean, I would applaud if this were directed at misusing error analysis and statistics, as I criticized before, but this seems to go against the media again, not the scientists.
The comic kind of insinuates that a 5% uncertain test needs only 20 repetitions (count the panels!) in order to have one false positive. The naïve calculation here is 20 * 5% = 100%. This is misleading of course. In fact, the percentage is only
1 - (1 - 0.05)20 = 1 - 0.9520= 64%
The comic is also wrong in that it goes on to test after already having found a positive result, here implicating that it is the media who afterward find out about it and blow it up. The scientific practice in POP science is largely to publish immediately after the spurious result (not at the LHC, but in most fields with smaller groups, say in nanotechnology for example, that is how it is done).
XKCD is often fun but sometimes it irks in terms of “look, science is all awesome, I understand science, so I am awesome”. All negative is to be blamed on the media and politics, but science is Jesus, so who won’t cheer for Jesus? If XKCD understood science deeply, it would produce a whole lot more jokes stabbing into a different direction, but maybe there is the fear to thereby lose some of the over enthusiastic audience.
So, ha ha ha, but with a bitter after taste nonetheless.