R/idMissingLevels.R

In mnodzenski/metabomxtr: A package to run mixture models for truncated metabolomics data with normal or lognormal distributions

idMissingLevels<-function(yname, missing.levels.list, dataset){

	##########################################################################################
	#This function determines the specific level of a categorical variable with completely 
	#missing outcome data, given that the categorical variable is known to be missing outcome 
	#data for at least one level.  
	#
	#Args:
	#	yname: metabolite column name (character string)
	#   missing.levels.list: a list output by function anyMissingLevels indicating whether any levels
    #						 of categorical variables in the mixture model have entirely missing 
	#						 metabolite values (and therefore cannot be modelled).  
	#	dataset: a data frame containing metabolite levels and categorical variables 
	#
	#Returns:
	#	A list containing the specific levels of model categorical variables with 
	#	no metabolite data 
	###########################################################################################

	#get the element of the missing levels list corresponding to yname
	target.list<-missing.levels.list[[yname]]
	
	#identify the variables with any missing levels 
	missing.level.vars<-names(target.list)[target.list]
	
	#loop over those variables and pull out the actual values of the missing levels 
	missing.level.vals<-lapply(missing.level.vars, function(missing.var){
	
		yname.vec<-dataset[ , yname]
		subset.data<-dataset[!is.na(yname.vec), missing.var, drop=F]
		subset.levels<-unique(subset.data[ , missing.var])
		original.levels<-unique(dataset[ , missing.var])
		missing.levels<-original.levels[! original.levels %in% subset.levels]
		return(missing.levels)
	
	
	})
	names(missing.level.vals)<-missing.level.vars
	return(missing.level.vals)

}