28 April 2015 by lberuti
In DataGrapple, the CDS Treemap Machine Tree presents data which grouped by statistical clusters. It groups together entities which have been historically close from a statistical point of view over a long period of time. It is quite striking to see that it is most of time reasonably easy to stick a name on a given cluster. Most associations make sense intuitively, and groups composed over long period of time still exhibit strong correlation over shorter periods. For instance, the brightest tones of red (which in this case means these are names which have seen their risk premia widen over the last week) are all concentrated on the right end side of the grapple, in 2 groups which have both been impacted by the poor statistics released last week regarding new home sales: home builders and monolines.