fit_loess: Fit RT Projection Model With LOESS

In hhabra/metabCombiner: Method for Combining LC-MS Metabolomics Feature Measurements

Fit RT Projection Model With LOESS

Description

Fits a local regression smoothing spline through a set of ordered pair retention times. modeling one set of retention times (rty) as a function on the other set (rtx). Filtering iterations of high residual points are performed first. Multiple acceptable values of span can be used, with one value selected through 10-fold cross validation.

Usage

fit_loess(
  object,
  useID = FALSE,
  spans = seq(0.2, 0.3, by = 0.02),
  outlier = c("MAD", "boxplot"),
  coef = 2,
  iterFilter = 2,
  prop = 0.5,
  weights = 1,
  rtx = c("min", "max"),
  rty = c("min", "max"),
  message = TRUE,
  control = loess.control(surface = "direct", iterations = 10)
)

Arguments

object	a metabCombiner object.
useID	logical. If set to TRUE, matched ID anchors detected from previous step will never be flagged outliers.
spans	numeric span values (between 0 & 1) used for loess fits
outlier	Thresholding method for outlier dection. If "MAD", the threshold is the mean absolute deviation (MAD) times coef; if "boxplot", the threshold is coef times IQR plus 3rd quartile of a model's absolute residual values.
coef	numeric (> 1) multiplier for determining thresholds for outliers (see outlier argument)
iterFilter	integer number of outlier filtering iterations to perform
prop	numeric. A point is excluded if deemed a residual in more than this proportion of fits. Must be between 0 & 1.
weights	Optional user supplied weights for each ordered pair. Must be of length equal to number of anchors (n) or a divisor of (n + 2)
rtx	ordered pair of endpoints for rtx; if "max" or "min", gives the maximum or minimum rtx, respectively, as model endpoints for rtx
rty	ordered pair of endpoints for rty; if "max" or "min", gives the maximum or minimum rtx, respectively, as model endpoints for rty
message	Option to print message indicating function progress
control	control parameters for loess fits; see: ?loess.control

Value

metabCombiner object with model slot updated to contain a fitted loess model

See Also

selectAnchors,fit_gam

Examples

data(plasma30)
data(plasma20)

p30 <- metabData(plasma30, samples = "CHEAR")
p20 <- metabData(plasma20, samples = "Red", rtmax = 17.25)
p.comb = metabCombiner(xdata = p30, ydata = p20, binGap = 0.0075)
p.comb = selectAnchors(p.comb, tolmz = 0.003, tolQ = 0.3, windy = 0.02)

#version 1
p.comb = fit_loess(p.comb, spans = seq(0.2,0.3,0.02), iterFilter = 1)

#version 2 (using weights)
anchors = getAnchors(p.comb)
weights = c(2, rep(1, nrow(anchors)), 2)  #weight = 2 to boundary points
p.comb = fit_loess(p.comb, spans = seq(0.2,0.3,0.02), weights = weights)

#version 3 (using identities)
p.comb = selectAnchors(p.comb, useID = TRUE, tolmz = 0.003)
p.comb = fit_loess(p.comb, spans = seq(0.2,0.3,0.02), useID = TRUE)

#to preview result of fit_loess
plot(p.comb, fit = "loess", xlab = "CHEAR Plasma (30 min)",
     ylab = "Red-Cross Plasma (20 min)", pch = 19,
     main = "Example fit_loess Result Fit")

hhabra/metabCombiner documentation built on Sept. 13, 2022, 6:25 a.m.