trim_outlier: Trim outlier out by Robust PCA

In Rundmus/BAf-R_package: Binder Array data hanlding package in facility

Description

Trim outlier out based on the orthogonal and score distances computed by robust principal components analysis (PCA). After log-transformation, like ordinary PCA, the values are scaled, but using robust statistics such as median instead of mean and MAD instead SD. Please note this function is applied to each binder batch separately and omits the data of the samples having any NA first.

Usage

trim_outlier(X, ...)

## S4 method for signature 'matrix'
trim_outlier(
  X,
  ...,
  alpha = 0.9,
  cutoff.od.P = 0.025,
  cutoff.sd.P = 0.025,
  coord = c("o&s", "o", "s", "os"),
  plotit = FALSE,
  main = "",
  kmax = 10
)

## S4 method for signature 'BAf'
trim_outlier(
  X,
  ...,
  by_s = NULL,
  alpha = 0.9,
  cutoff.od.P = 0.025,
  cutoff.sd.P = 0.025,
  coord = c("o&s", "o", "s", "os"),
  plotit = FALSE,
  kmax = 10
)

Arguments

X	a matrix or a BAf-class object
...	in order to encourage to use all parameter names
alpha, kmax	the parameter for rrcov::PcaHubert
cutoff.od.P, cutoff.sd.P	the probability threshold for the orthogonal and score distances
coord	the distance coordinates to be used in outlier classification. "o&s" indicates the points beyond the cutoffs on both coordinates are labelled as outliers. In other options, the points over any cutoff on any coordinates are marked as outliers.
plotit	if plots were wanted
main	title of plot
by_s	Robust PCA per sample set divided by this. If it is a character, then the column named as it in @sinfo wil be used for stratification. When a factor vector is given, then it is used as it is in dividing into groups. If it is NULL as the default, there will be no stratification.

Value

The BAf object after outlier removal

Author(s)

Mun-Gwan Hong <[e-mail]>

References

Hubert, M., Rousseeuw, P. J., Branden, K. V., (2005) ROBPCA: A New Approach to Robust Principal Component Analysis. Technometrics 47, 64-79

See Also

apply_per_group

Examples

data(sba)
B <- trim_outlier(sba, applyBy= "plate", plotit = FALSE)
summary(B)

Rundmus/BAf-R_package documentation built on May 18, 2020, 12:59 p.m.