“Interesting” Statistical Techniques for Archival SETI

In Annis (1999), an archival data based search is conducted to see if any nearby galaxies display the characteristics of hosting an extremely advanced ETI. In particular, the paper looks for Kardashev type III civilization (one that can control energy on the scale of its entire host galaxy) that is converting most of the starlight in its galaxy into usable energy. Deflecting large amounts of starlight should noticeably dim the surface brightness of the galaxy and would show up clearly as an outlier in the well-documented galaxy relations. The author searches for these outliers in the brightness and temperature data from spiral galaxies in clusters and nearby elliptical galaxies. None are found.

One question I had while reading this paper was how would we know a dim galaxy were there? I mean, if almost all of the starlight is being blocked, then how would we even know to look at that place in the sky? Presumably, we would see a dim galaxy’s gravitational effects on its neighbors, which may have been why the study focusses somewhat on galaxies in nearby spiral galaxy clusters (Virgo and Ursa Major).

The author chooses a pretty arbitrary floor for what he thinks is an outlier. He decides that any galaxy that is 75% dimmer than its respective (spiral or elliptical galaxy) relation is an outlier. His methodology seems pretty circular though. He visually finds what he thinks is where, if one assumes there is Gaussian scatter about the relations, a cutoff is that is “at some high sigma enough to avoid statistical outliers”. There are certainly more robust ways to calculate outliers (including finding the Mahalanobis distances from each data point to the sample means).

An interesting calculation that is performed at the end of the article constrains the emergence time of a star-fed Kardashev type III civilization. Assuming that there is a constant chance of a KIII civ occurring throughout time, you can use Poissonian statistics and the fact that since it appears that all of the surveyed galaxies (including our own) have no KIII civilizations to put a lower limit on the occurrence time for these civilizations. Each new galaxy that they aren’t detected in adds ~10 Gyr of non-detection time. The paper arrives at a lower limit of ~6.5 billion years at 99% confidence. Neat.