metaDMG-dev / metaDMG-cpp

metaDMG-cpp
4 stars 2 forks source link

dfit aggregated results #16

Open fgvieira opened 11 months ago

fgvieira commented 11 months ago

When the option --nodes is specified, dfit aggregates the results. Don't think it makes much sense to be here, since dfit should just optimize the matrices it gets as input.

On the other hand, if LCA bdamage prints matrices for non-specific levels (see #14), these might be being counted twice.

ANGSD commented 11 months ago

I do not have a problem with the internal mismatch matrices being in memory solely. Is it because you want to observe the internal mismatch matrices?