imputeRowMinRand: Impute missing values with random numbers based on the row...

In xiaodfeng/DynamicXCMS: LC-MS and GC-MS Data Analysis

Impute missing values with random numbers based on the row minimum

Description

Replace missing values with random numbers from a normal distribution based on (a fraction of) the row min and a standard deviation estimated from the linear relationship between row standard deviation and mean of the full data set. Parameter sd_fraction allows to further reduce the estimated standard deviation.

Usage

imputeRowMinRand(x, min_fraction = 1/2, sd_fraction = 1, abs = TRUE)

Arguments

x	matrix with abundances, rows being features/metabolites and columns samples.
min_fraction	numeric(1) with the fraction of the row minimum that should be used to replace NA values in that row.
sd_fraction	numeric(1) factor to reduce the estimated standard deviation.
abs	logical(1) to force imputed values to be strictly positive.

Details

Imputed values are taken from a normal distribution with mean being a user defined fraction of the row minimum and the standard deviation estimated for that mean based on the linear relationship between row standard deviations and row means in the full matrix x.

To largely avoid imputed values being negative or larger than the real values, the standard deviation for the random number generation is estimated ignoring the intercept of the linear model estimating the relationship between standard deviation and mean. If abs = TRUE NA values are replaced with the absolute value of the random values.

Author(s)

Johannes Rainer

See Also

imputeLCMD package for more left censored imputation functions.

Other imputation functions: imputeRowMin

Examples

library(faahKO)
data("faahko")

xset <- group(faahko)
mat <- groupval(xset, value = "into")

## Estimate the relationship between row sd and mean. The standard deviation
## of the random distribution is estimated on this relationship.
mns <- rowMeans(mat, na.rm = TRUE)
sds <- apply(mat, MARGIN = 1, sd, na.rm = TRUE)
plot(mns, sds)
abline(lm(sds ~ mns))

mat_imp <- imputeRowMinRand(mat)

head(mat)
head(mat_imp)

xiaodfeng/DynamicXCMS documentation built on Aug. 6, 2023, 3:02 p.m.