Saturday, April 30, 2016

Predictions with a single observation

A recent study from the University of Rochester (1) claims to improve the "Drake equation," - a constant reminder that multiplying random numbers does not provide any additional insights that are not present in the components. Now with exoplanets galore, the study that appeared in Astrobilology claims they can put a "pessimistic estimate" on the probability of non-existence of advanced civilizations, elsewhere in the universe. Just as in the previous attempt, the study suffers from traditional statistics. Most agree that a singular observation is not sufficient to predict anything, let alone the age, capabilities and the distance from Earth, where they could reasonably exist.

As the authors note, the space time window afforded to humans is too narrow to prove or disprove anything. However, the tendency to be amazed by large numbers and the multiplicative effects of such constructs, have led scientists to make non-scientific claims. Until at least a second data point becomes available, the effort expended on statistical analysis in this area is a waste of time. Availability of data is necessary but not sufficient to assign probabilities. Even those clinging to normality statistics, centuries old by now, know that it is not a good tool to make predictions.

More importantly, those awaiting ET's arrival, have almost infinite flexibility to keep on searching. If one has a hypothesis, then an accumulation of negative findings against it, regardless of how many trials are possible, has to be given due consideration. As an example, if one claims favorable conditions exist for life on Enceladus, Saturn's famous moon, such as water, oxygen and a heat source - then investing into the exploration of the icy rock is reasonable. However, if one comes out empty, it cannot be irrelevant. Just because there are trillion other rocks, in the solar system alone, that could be explored, one cannot simply ignore such an observation. At the very least, it should challenge the assumptions used by the space agency and others to justify such explorations. This "new" statistics - perhaps called "Statistics of large numbers," - where no negative observation has any utility - is very costly even though it is well positioned to pump out publications.

Scientists, engaged in irrelevant and invalid observations. aided by large numbers, may need to challenge themselves to advance the field.