
March 14, 2023
	in deviation() add pearson residuals
	need to pass in y= vobj


Feb 28
DONE: Zenith on mvTest results


April 19, 2021
	Add treat and topTreat

March 30, 2021
	check dream + eBayes with trend and robust


#Authors@R: c(person("Gabriel", "Hoffman", role = c("aut", "cre"), email = "gabriel.hoffman@mssm.edu"))


DONE!
March 18, 2021
	info must be casted to a data.frame since DataFrame is not handled correctly
		and gives error:
		Error in if (inherits(possibleError, "error") && grep("the fixed-effects model matrix is column rank deficient",  : 
	  missing value where TRUE/FALSE needed

	droplevels in dream


# November 2, 2020
- integrate voomWithQualityWeights() with voomWithDreamWeights
	o Need to convert lm.fit() calls to dream()
	o Need to dream() to return hatvalues



recommend placing a check to verify if the names in the info matrix, match with the names used in the formula and return an error if not




library(BiocParallel)

# globally specify that all multithreading using bpiterate from BiocParallel
# should use 8 cores
register(SnowParam(8))

# By default it is set to the max number of CPUs
# If you have a very long interactive session, this can crash and you get the error you see

Alternatively, you can specify and restart the parallel backend every function call


fitExtractVarPartModel(..., BPPARAM=SnowParam(8))

fitVarPartModel(..., BPPARAM=SnowParam(8))

dream(..., BPPARAM=SnowParam(8))



# July 18, 2018
Check if there is variation in each gene.  
	else throw error


# June 16, 2016
Improve warnings in fitVarPart and fitExtractVarPart to show which gene give a warning

# April 16, 2016
 Make correlation matrix and linear mixed model equantions in vignette
 the same as in supplement.  The vignette examples are wrong

Add canCorPairs to the vignette


# Feb 18, 2015
add e_{i,k} in ICC correlation equation


# December 14, 2015

# Add strong warning to fitVarPartModel() that estimates memory usage.  In vignette, users should be discoraged from using this interface.

# use refit() to fit the model for each gene:
#  cannot do this because refit() doesn't accept new weights

# DE test: show increased power when correlation is overestimated, and decrease false positive rate when correlation is underestimated.  (order is correct?)  





# adjust doesn't work with varying coefficient model, but it throwings an error now

# run simulations with varying coefficient model