{optmatch}
as a suggested package since it is
no longer on CRAN.LazyData: true
from DESCRIPTION since there is
no ‘data’ directory.Fixes a lot of things for CRAN resubmission.
seqgendiff::EigenDiff()
. Replaces its usage
with cate::est.factor.num()
. This is fine since it was only
used in the now defunct seqgendiff::poisthin()
.{optmatch}
package
is now only suggested rather than imported. This is because the
{optmatch}
package is under a super weird license that I
didn’t previously know about.{clue}
package, seems to work just as well as {optmatch}
, and so I
added it as an option. However, since I used {optmatch}
in
the simulations for the paper, I have kept
permute_method = "optmatch"
as the default option.select_counts()
, a function that will subsample
the rows (genes) and columns (samples) of a RNA-seq count matrix. It is
generally recommended that you do this subsampling each iteration of a
simulation study so that your results do not depend on the specific
structure of your data. The samples are just selected randomly. There
are four different criteria for selecting the genes.thin_all()
, a function that uniformly thins all
counts.This has been a massive rewrite of the {seqgendiff}
package.
poisthin()
is now defunct. The two-group model is now
implemented in the thin_2group()
function. I’ll keep it
around since some of my old simulation code depends on it.thin_diff()
,
thin_2group()
, thin_lib()
, and
thin_gene()
.poisthin()
, do not
have functionality to subset count matrices. This is on purpose. I
wanted the functionality of these thinning functions to be simpler.poisthin()
, which can only handle the two-group
model, thin_diff()
can handle generically any design, while
still controlling the level of correlation between the design variables
and the surrogate variables.ThinDataToSummarizedExperiment()
and
ThinDataToDESeqDataSet()
.corassign()
lets you make group assignment that is
correlated with hidden factors.poisthin()
, the group_assign = "cor"
option uses corassign()
to make group assignments.