Another kind of astronomical objects has been successfully analyzed with maximum parsimony: Gamma Ray Bursts!
These are still mysterious objects, extremely violent, that emit a lot of gamma-rays during tenths of a second to several seconds. Gamma-rays are very energetic electromagnetic radiation, more so than X-rays, UV and of course visible light. Gamma-ray detectors on satellites cannot provide an acute position on the sky, only the visible telescopes can. But within tenths of a second you must detect the event, send an alert to an X-ray satellite that can point the object and narrow its position for a visible telescope to take on the characterization of the event. Only in these conditions have we been able to know that they are cosmological, that is distributed in galaxies throughout the entire Universe, and the information in visible light can sometimes provide its distance (indeed its redshift).
Since they are very bright, they can be seen very very far in the Universe, much farther away than our farthest candles up to now, the Supernovae Ia that have probed the geometry of the Universe and recently where the subject of a Nobel Prize. But Gamma-Ray Bursts can be seen much much farther away and thus probe a huge volume of the Universe.
The idea of this work, suggested by my collaborator Vincenzo Cardone, is to repeat the history. Each time astronomers used a candle to determine the distance scale of the Universe, they were confronted to several families for each kind of objects. This started with the determination of the Hubble constant which proved wrong when it was realized that there are several families of the variable stars used for this purpose, the Cepheids. Similarly, a scaling relation appears only when the Supernovae of type Ia are isolated.
Since there seems to exist a scaling relation for Gamma-Ray Bursts that potentially can be of same interest as the Supernova but reaching much higher redshifts, it is necessary to ensure that the sample of objects used form an homogeneous family.
Our result is that there are nine groups, four of which are more robust. The scaling relation is different within each family than globally, and may even disappear. This could be a very important result, but unfortunately, the observational uncertainties are high and the number of objects still low. We cannot provide a firm conclusion, but more investigations will be following, for sure!
Post Scriptum: Well, why cladistics should work on such objects? Where is the transmission with modification process? Where is evolution? This will be the subjects of other posts since we have already one paper ready that should answer these questions. And I probably already wrote that I am more and more convinced that cladistics belongs to graphical techniques of clustering, especially for continuous and quantitative data. The proof is that it works, it often agrees with kmeans (or kmedoids in the present paper) partitioning, and it provides physically plausible interpretations.