seqgendiff 1.2.2

seqgendiff 1.1.1

Updates to the documentation. I included citations throughout the documentation to direct users to where they can get details on the method.
The citation file now indicates the preprint on bioRxiv.

Fixes a lot of things for CRAN resubmission.

Adds more information on return values.
Changes title to less than 65 characters.
Uses local environments rather than global environment to assess messages.
Removes seqgendiff::EigenDiff(). Replaces its usage with cate::est.factor.num(). This is fine since it was only used in the now defunct seqgendiff::poisthin().

The biggest change here is that the 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 https://cran.r-project.org/package=optmatch/LICENSE.
The user may also now specify the permutation method in the thinner functions.
The Hungarian algorithm, implemented in the 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.

I added 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.
I also added citation information. This will be updated after I submit the manuscript to bioRxiv.

This version mostly updates the documentation.
Beyond this major documentation update, we also added 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.
The main functions for thinning are now thin_diff(), thin_2group(), thin_lib(), and thin_gene().
Note that these functions, unlike poisthin(), do not have functionality to subset count matrices. This is on purpose. I wanted the functionality of these thinning functions to be simpler.
Unlike 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.
Functions now exist to convert the simulation output to common Bioconductor S4 classes. See ThinDataToSummarizedExperiment() and ThinDataToDESeqDataSet().

corassign() lets you make group assignment that is correlated with hidden factors.
In poisthin(), the group_assign = "cor" option uses corassign() to make group assignments.