summary.agraph_fit: Summary for the fitted graph.

In mailund/admixture_graph: Admixture Graph Manipulation and Fitting

Description

Prints:
Optimal admixture proportions and a complaint if some of them are not trurly fitted, i. e. if after fixing a (possibly empty) subset of them, the rest have typically no effect on the cost function. Here typically means that some isolated values of the admixture proportions, like 0 or 1, might actually give a significantly worse fit than the constant fit given by any other values (but not better). Thus, an admixture proportion not being fitted does not always mean that there is no evidence of an admix event, as fixing them at 0 or 1 could make the fit worse while the exact value won't matter otherwise.
The optimal edge lengths give one of the solutions for the best fit. It is generally not unique, as after fixing the admixture proportions, the best edge lengths are a non-negative least square solution for a system of linear equations. To get all the solutions one has to add any solution of the corresponding homogeneous system to the given exaple solution (and exclude possible negative values). The solutions of the homogeneous system are given as a set of free edge lengths that may obtain any non-negative value, and bounded edge lengths that linearly depend on the free ones.
Minimal error is the value of the cost_function at the fit.

Usage

## S3 method for class 'agraph_fit'
summary(object, ...)

Arguments

object	The fitted object.
...	Additional parameters.

mailund/admixture_graph documentation built on May 21, 2019, 11:06 a.m.