to_do
Suggestions/fixes for blupf90 suite
Things that make user's life difficult
idsol
is postnumf90
but is not really well documented
DONE: Open all files with status=old
to avoid creating new empty file (cryptic error on output)
Plain text readable file gibbs_samples
with colnames R11 R12 … G11_11 G12_11…
(postgibbsf90 is not easy and graphs do not work in most computers)
DONE: use get_command_argument()
in order not to use echo file.par | blupf90
. Now one can use blupf90 file.par
for all software
DONE for gibbs1f90, gibbs2f90, gibbs3f90 : use getarg
to pass gibbs arguments: gibbs2f90 mygibbs.par –rounds 10000 –burnin 1000 –thin 10 –thinprint 20
DONE: informative help in inbupgf90
: sends to wiki
DONE: use method 2 (in memory) as default in inbupgf90
: done
in several applications UPGs are conceived to come at the end of the renumbered pedigree. This is undocumented.
Suggestions
Do A-1 with consideration of inbreeding unless requested not to
OPTION skip datafile n
to skip n lines (possibly colnames)
OPTION skip pedfile n
to skip
DONE for gibbs1f90, gibbs2f90, gibbs3f90 : allow all missing in gibbs*f90
(in order not to edit files !!)
Bigger things
Options for sampling variances for GWAS. Current option gives pretty good Manhattan plots, But accuracies are improved with large but not small QTLs, and the iteration diverges. New options could mimic other methods where all SNP effects are sampled simultaneously, such as NonlinearA or FastBayesA.
-
Derenumbering program
to_do.txt · Last modified: 2017/07/07 01:01 by ignacy