examine_edge_optimisation_matrix: Examine the edge optimisation matrix to detect unfitted admix variables.
Description
If the essential number of equations is not higher than the essential number of
edge variables, the quality of edge optimisation will not depend on the admix
variables (expect possibly in isolated special cases where the quality can be worse),
and a complaint will be given.
Note: The admix variable not being fitted does not mean that there is no evidence of
an admix event! Isolated values of the admix variables, possibly $0$ or $1$,
might give significantly worse fit than a typical value (but not the other way around).
Usage
examine_edge_optimisation_matrix(matrix, tol = 1e-08)
Arguments
tol
Calulating the rank with qr.solve
sometimes crashes.
Default $10^(-8)$. Value
An indicator of warning (complaint
), coding all the possibilities in
a way that is interpreted elsewhere (in summary.agraph_fit
).