closely related to #24 (in that that can be built on top of this): we should be able to produce some representation of the tree with each node annotated with its error - not the sum of its leaves' errors, but the error of the sums of its leaves' distributions.
closely related to #24 (in that that can be built on top of this): we should be able to produce some representation of the tree with each node annotated with its error - not the sum of its leaves' errors, but the error of the sums of its leaves' distributions.