replNAbyLow: Replace NAs by low values

In wrMisc: Analyze Experimental High-Throughput (Omics) Data

Replace NAs by low values

Description

With several screening techniques used in hight-throughput biology values at/below detection limit are returned as NA. However, the resultant NA-values may be difficult to analyse properly, simply ignoring NA-values mat not be a good choice. When (technical) replicate measurements are available, one can look for cases where one gave an NA while the other did not with the aim of investigating such 'NA-neighbours'. replNAbyLow locates and replaces NA values by (random) values from same line & same group 'grp'. The origin of NAs should be predominantly absence of measure (quantitation) due to signal below limit of detection and not saturation at upper detection limit or other technical problems. Note, this approach may be not optimal if the number of NA-neighbours is very low. Replacamet is done -depending on agrument 'unif'- by Gaussian random model based on neighbour values (within same group), using their means and sd, or a uniform random model (min and max of neighbour values) . Then numeric matrix (same dim as 'x') with NA replaced is returned.

Usage

replNAbyLow(
  x,
  grp,
  quant = 0.8,
  signific = 3,
  unif = TRUE,
  absOnly = FALSE,
  seed = NULL,
  silent = FALSE,
  callFrom = NULL
)

Arguments

x	(numeric matrix or data.frame) main input
grp	(factor) to organize replicate columns of (x)
quant	(numeric) quantile form 'neighbour' values to use as upper limit for random values
signific	number of signif digits for random values
unif	(logical) toggle between uniform and Gaussian random values
absOnly	(logical) if TRUE, make negative NA-replacment values positive as absolute values
seed	(integer) for use with set.seed for reproducible output
silent	(logical) suppress messages
callFrom	(character) allow easier tracking of message(s) produced

Value

numeric matrix (same dim as 'x') with NA replaced

See Also

naOmit, na.fail

Examples

dat <- matrix(round(rnorm(30),2),ncol=6); grD <- gl(2,3)
dat[sort(sample(1:30,9,repl=FALSE))] <- NA
dat; replNAbyLow(dat,gr=grD)

wrMisc documentation built on Nov. 17, 2023, 5:09 p.m.