OutlierDM (version 1.1.1)

odm: Outlier Dectection for Multi-replicated data

Description

This function provides some routines for detecting outlying observations (peptides) for multi-replicated high-throughput data, especially in LC/MS experiments.

Usage

odm(x, k = 3,  quantreg = c("linear", "nonlin", "constant", "nonpar"),  method = c("proj", "diff", "pair", "grubbs", "dixon",  "iqr",  "siqr", "Zscore"), ...)

Arguments

x

data vectors or matrices. These can be given as named arguments. If the number of predictors is 2, x1 describes one n-by-1 vector for data and x2 describes the other n-by-1 vector for data (n= number of peptides, proteins, or genes)

k

non-negative tuning parameter for the outlier detection algorithm. For IQR-based algorithms such as 'iqr', 'siqr', 'proj', 'diff', and 'pair', it works in the formula of Q1-k*IQR and Q3+k*IQR, where IQR=Q3-Q1. For 'Zscore', it works for the 'k' in |Z| > k. A default value is 3.

quantreg

type of quantile regression models used for the outlier detection method. You can use one of the 'constant', 'linear', 'nonlin', and 'nonpar' which mean the constant, linear, non-linear, and non-parametric quantile regression in order. For more details, see the quantreg package.

method

type of outlier detection methods. You can select one of the 'Zscore', 'iqr', 'dixon', 'grubbs', 'pair', 'diff', and 'proj' algorithms as follows. Zscore: Z-score based criterion (Cho and Eo, 2015) iqr: Interquartile range (IQR) criterion (Cho and Eo, 2015) siqr: Semi-interquartile range (IQR) criterion (Cho and Eo, 2015) dixon: Dixon's test (Dixon, 1950; 1951) grubbs: Grubbs test (Grubbs, 1950; 1969) pair: Pariwise OutlierD algorithm (Cho et al., 2008; Eo et al., 2012) proj: Projection-based OutlierD algortihm (Eo et al., 2012) diff: Difference-based OutlierD algorithm (Eo and Cho, 2015)

...

minor tuning parameters used in odm.control(). See odm.control.

Value

call:

evaluated function call

raw.data:

raw dataset used in the model fitting

res:

result matrix of the model fitting. It consists of used data set with some transformation and outlying statistic.

x.pair:

Object of class "list"

k:

threshold parameter for constructing outlier detection methods

outlier:

matrix including the status of each outlying peptide and sample

n.outliers:

the number of outlying parameters (peptides) to be detected by the model fitting.

quantreg:

type of quantile regression used for the model fitting

method:

type of outlier detection method used for the model fitting

contrl.para:

a list of minor parameters

References

Eo, S-H and Cho, H (2015) OutlierDM: More robust outlier detection algorithms for multi-replicated high-throughput data.

Cho, H and Eo, S-H. (2015) Outlier detection for mass-spectrometry data.

Eo, S-H, Pak D, Choi J, Cho H (2012) Outlier detection using projection quantile regression for mass spectrometry data with low replication. BMC Res Notes.

Cho H, Lee JW, Kim Y-J, et al. (2008) OutlierD: an R package for outlier detection using quantile regression on mass spectrometry data. Bioinformatics 24:882--884.

Grubbs FE (1969) Procedures for detecting outlying observations in samples. Technometrics 11:1--21.

Dixon WJ (1951) Ratios involving extreme values. Ann Math Statistics 22:68--78.

Dixon WJ (1950) Analysis of extreme values. Ann Math Statistics 21:488--506.

Grubbs FE (1950) Sample criteria for testing outlying observations. Ann Math Statistics 21:27--58.

See Also

OutlierDM-package to provide the general information about the OutlierDC package OutlierDM-class to provide the information about the "OutlierDM" class odm.control to control tuning parameters

Examples

  ## Not run: 
#     ##############################################################
#     #
#     #   Outlier Detection for Mass Spectrometry Data
#     #   Section 3. Illustration
#     #   by HyungJun Cho and Soo-Heang Eo,
#     #   Dept of Statistics, Korea University, Seoul, Korea
#     #
#     ##############################################################
# 
#     #####
#     # Load a package OutlierDM
# 
#     # If an OutlierDM package is not installed on your system, type
#     #install.package('OutlierDM', dependency = TRUE)
#     library(OutlierDM)
# 
#     #####
#     # Sec 3.1 When the number of replicates is large enough
#     ## Load toy dataset
#     data(toy)
#     head(toy)
#     pairs(log2(toy), pch = 20, cex = .7)
# 
#     #####
#     # Fit 1. Z-score based criterion
#     fit1 = odm(x = toy, method = "Zscore", k = 3)
#     fit1
#     summary(fit1)
#     head(input(fit1))
#     head(output(fit1))
#     print(outliers(fit1), digits = 3)
#     plot(fit1)
#     rect(1, -4, 10, 4, col = heat.colors(20,alpha = 0.3), border = heat.colors(20,alpha = 0.5))
#     oneplot(object = fit1, i = 4)
#     title("Outlier Detection by the Z-score criterion")
# 
#     # Add a peptide name on a dot-plot
#     #oneplot(fit1, 191,1)
#     #title("Outlier Detection by the Z-score criterion")
# 
#     #####
#     # Fit 2. Grubbs test criteria
#     fit2 = odm(x = toy, method ="grubbs", alpha = 0.01)
#     fit2
#     summary(fit2)
#     head(output(fit2))
#     print(outliers(fit2), digits = 3)
#     oneplot(object = fit2, i = 1)
#     title("Outlier Detection by the Grubbs criterion")
# 
#     # Add text
#     #oneplot(fit2, 191,1)
#     #title("Outlier Detection by the Grubbs criterion")
# 
#     #####
#     # Fit 3. IQR criteria
#     fit3 = odm(x = toy, method = "iqr", k = 3)
#     fit3
#     summary(fit3)
#     print(outliers(fit3), digits = 3)
#     plot(fit3)
#     rect(1, -4, 10, 40, col = heat.colors(20,alpha = 0.3), border = heat.colors(20,alpha = 0.5))
#     oneplot(fit3, 1)
#     title("Outlier Detection by the IQR criterion")
# 
#     # Add a peptide name on a dot-plot
#     #oneplot(fit3, 1, 1)
#     #title("Outlier Detection by the IQR criterion")
# 
#     #####
#     # Fit 4. SIQR criteria
#     fit4 = odm(x = toy, method = "siqr", k = 3)
#     fit4
#     summary(fit4)
#     print(outliers(fit4), digits = 3)
#     plot(fit4)
#     rect(1, -4, 10, 4, col = heat.colors(20,alpha = 0.3), border = heat.colors(20,alpha = 0.5))
#     oneplot(fit4, 1)
#     title("Outlier Detection by the SIQR criterion")
# 
#     #####################
#     ## Real data example
#     #####################
#     data(lcms3)
#     head(lcms3)
#     pairs(log2(lcms3), pch = 20, cex = .7)
# 
#     #####
#     # Fit 5. OutlierD
#     fit5 = odm(lcms3[,1:2], method = "pair", k = 3)
#     fit5
#     summary(fit5)
#     head(output(fit5))
#     print(outliers(fit5), digits = 3)
#     plot(fit5)
#     title("Outlier Detection by the OutlierD algorithm")
# 
#     #####
#     # Fit 6. OutlierDM 
#     fit6 = odm(lcms3, method = "proj", k = 3, center = TRUE)
#     fit6
#     summary(fit6)
#     print(outliers(fit6), digits = 3)
#     plot(fit6)
#     title("Outlier Detection by the OutlierDM algorithm")
#     oneplot(fit6, 18)
#     #oneplot(fit6, 18, 1)
#     title("The dotplot for the 18th samples of the lcms3 data")
#     ### End of the illustration
# 
#     
#     #####
#     # Other OutlierDM algorithms
#     data(lcms3)
#     
#     ## Load 
#     ## Fit projection approaches
#     fit.proj.const <- odm(lcms3, quantreg="constant")
#     fit.proj.linear <- odm(lcms3, quantreg="linear")
#     fit.proj.nonlin <- odm(lcms3, quantreg="nonlin")
#     fit.proj.nonpara <- odm(lcms3, quantreg="nonpar", lbda = 1)
#     
#     par(mfrow = c(2,2))
#     plot(fit.proj.const, main = "Constant")
#     plot(fit.proj.linear, main = "Linear")
#     plot(fit.proj.nonlin, main = "NonLinear")
#     plot(fit.proj.nonpara, main = "Nonparametric")
# 
#   ## End(Not run)