MetabolomicsBasics (version 1.4.5)

RemoveFactorsByANOVA: RemoveFactorsByANOVA.

Description

RemoveFactorsByANOVA will remove variance from data using an ANOVA model.

Usage

RemoveFactorsByANOVA(
  y = NULL,
  sam = NULL,
  fmod = NULL,
  kmod = NULL,
  output = c("y_norm", "y_lm", "anova_y", "anova_y_norm", "boxplot")[1],
  remove_outliers = 0
)

Value

Depends on output. Usually the normalized data vector (or matrix).

Arguments

y

Data vector (or data matrix) to normalize (numeric + in same order as sam).

sam

data.frame containing the factors or numerical vars for ANOVA model.

fmod

Full model describing the experimental setting (provided as character string).

kmod

Reduced model describing all the biological factors to keep (provided as character string).

output

Should be y_norm in general but can be switched for testing.

remove_outliers

Should be a numeric integer x (with $x=0$ : no effect; $x>=1$ remove all values which have error e with $e > abs(mean + x * sd)$ ).

Details

See examples.

Examples

# set up sample information
sam <- data.frame(
  "GT" = gl(4, 10),
  "TR" = rep(gl(2, 5), 4),
  "Batch" = sample(gl(2, 20)),
  "Order" = sample(seq(-1, 1, length.out = 40))
)

# set up artificial measurement data
set.seed(1)
# specify main effects
m1 <- c(5, 6, 2, 9)[sam$GT] + c(-2, 2)[sam$TR]
m2 <- c(5, -6, 2, 4)[sam$GT] + c(-2, 2)[sam$TR]
# add run order bias and noise
m1 <- m1 + c(-3, 3)[sam$Batch] + 3 * sam$Order + rnorm(nrow(sam), sd = 0.5)
m2 <- m2 - 5 * sam$Order + rnorm(nrow(sam), sd = 0.8)
dat <- data.frame(m1, m2)

# apply function to remove variance
# full model incorporating all relevant factors defined in sample table
fmod <- "GT*TR+Batch+Order"
# reduced model: factors to be kept from full model; everything elso will be removed from the data
kmod <- "GT*TR"
RemoveFactorsByANOVA(y = dat[, "m1"], sam = sam, fmod = fmod, kmod = kmod, output = "anova_y")
RemoveFactorsByANOVA(y = dat[, "m1"], sam = sam, fmod = fmod, kmod = kmod, output = "anova_y_norm")