Abstract:
Observing and storing the photons of the incoming radiation from
the Cosmos typically give a data cube defined by (α, δ, λ).
It is easy to translate this data structure
into the formalism of multivariate statistics. A common problem in
the multivariate statistics is whether the stochastic variables
described by observed properties are statistically independent or
can be described by a less number of hidden variables. This is the
task of factor analysis. Forming groups from cases having similar
properties according to the measures of similarities or the
distances is the task of cluster analysis. We demonstrated in
three cases how these technics can be used for separating
physically independent cosmic components projected onto the
same celestial area by chance.
