imputationTest: Test for the various methods

In AurelieGuilbault/VIQCing: Data Processing For Metabolomic Data

Description

Test imputation accuracy for the given dataset with different method options. Compute the NRMSE values for the desired number of test. If asked, produces an output <filename>_Accuracy.txt with the columns

• "Method";

• "missing_proportion";

• "transformation";

• and "NRMSE"

Usage

imputationTest(input, output = NULL, k = 2, method = "knn",
  npcs = 3, sigma = 0.1, nbTest = 10, nTree = 30,
  na.string = "NA", missing = 0.05, missingType = "MCAR",
  transformation = "None", sampleStart = 3)

Arguments

input	file containing the test dataset; Should contain: all the columns sample from <sampleStart> to the end of the file; *MUST BE COMPLETE (no NA values) for accurate results
output	default NULL, name of the test results file if not NULL
k	default 2, the k used for the knn imputation;
method	default "knn", the chosen method for replacing the missing values. Can be "knn", "RF", "QRILC", "SVD", "mean", "median", "HM" or "0". See Details.
npcs	default 3, npcs for SVD method;
sigma	default 0.1, tune sigma parameter for QRILC method;
nTree	default 30, number of tree for the RF method;
na.string	default "NA", string to consider as NA in the dataset;
missing	default 0.05, proportion of missing values;
missingType,	default "MCAR" (missing completely at random), can be "MNAR" (not at ramdom), will target the values under the median;
transformation	default "None", can be "scale" or "log";
sampleStart	default 3, 1st column of the actual data;
nTest,	default 10, number of test to loop;
compound	default NULL, position of the compound column if named otherwise;
metabolite	default NULL, position of the metabolite column if named otherwise;

Details

Will compute de NRMSE (Normalized Root Mean Squared Error) for an imputation test. Available methods:

• "knn": From the impute package, use the k nearest neighboors to impute the values;

• "RF": From the missForest package, use RandomForest algorithm to impute the values;

• "QRILC": From the imputeLCMD package, use Quantile regression to impute the values;

• "SVD": From the pcaMethods package, use SVDimpute algorithm as proposed by Troyanskaya et al, 2001. to impute the values;

• "mean","median", ""median", "0", "HM": simple value replacement, either by the mean, median, 0 of Half minimum of the row;

Value

resDf, the result NRMSE dataframe.

See Also

• impute packagehttps://www.rdocumentation.org/packages/impute

• missForest package https://www.rdocumentation.org/packages/missForest

• imputeLCMD package https://www.rdocumentation.org/packages/imputeLCMD

• pcaMethods package https://www.rdocumentation.org/packages/pcaMethods

Examples

for a dataset with the following header ; Compound, m/z, Metabolite, RT, Sample #1, ...
imputationTest("dummySet.tsv", method="knn", transformation="log", sampleStart=5)

for a dataset with the following header ; compound, m/z, metabolite, RT, Sample #1, ...
imputationTest("dummySet.tsv", method="knn", transformation="log", metabolite = 3, compound = 1, sampleStart=5)

AurelieGuilbault/VIQCing documentation built on Oct. 30, 2019, 5:02 a.m.