MIMFA: Handling Missing Individuals in MFA

In GonzalezIgnacio/HandlingMissRows: Handling Missing Individuals in Multi-Omics Data Integration

Description

The MIMFA function estimates coordinates of individuals and variables on the MFA components by implementing a multiple imputation (MI) approach in order to deal with multiple tables in presence of missing individuals.

Usage

MIMFA(object, ncomp=2, M=NULL, estimeNC=FALSE, maxIter=500, tol=1e-10)

Arguments

object	an object of class MIDTList.
ncomp	a number of components to include in MFA when estimeNC=FALSE (default to 2). If estimeNC=TRUE, then ncomp correspond to the maximum number of components to test.
M	integer, number of imputations. Default to min(30, Mtotal), where Mtotal is the total number of possible imputations.
estimeNC	logical. If TRUE the number of MFA components for data imputation is estimated. Default is FALSE.
maxIter	integer, maximum number of iterations for the imputeDataMFA function.
tol	positive value, the threshold for assessing convergence in the imputeDataMFA algorithm.

Details

According to the MI methodology, missing individuals are filled in by several sets of plausible values, resulting in M completed data. MFA is then applied to each completed data leading to M different configurations. Finally, the M configurations are combined using the STATIS method to yield one consensus solution.

If estimeNC=TRUE, the number of MFA components for data imputation is estimated using the generalized cross-validation approximation method. In this case, ncomp corresponds to the maximum number of components to test.

Value

A MIDTList object containing additional slots for:

compromise

configurations

imputedIndv

MIparam

See MIDTList for description.

Author(s)

Ignacio González and Valentin Voillet

References

Voillet V., Besse P., Liaubet L., San Cristobal M., González I. (2016). Handling missing rows in multi-omics data integration: Multiple Imputation in Multiple Factor Analysis framework. BMC Bioinformatics, 17(40).

Lavit C., Escoufier Y., Sabatier R., Traissac P. (1994). The ACT (STATIS method). Computational Statistics & Data Analysis, 18(1), 97–119.

Josse J., Husson F. (2012). Selecting the number of components in PCA using cross-validation approximations. Computational Statistics and Data Analysis, 56, 1869–1879.

See Also

plotInd, plotVar, tuneM

Examples

#-- load data and create MIDTList object
data(NCI60)
midt <- MIDTList(NCI60$mae)
midt

#-- performs MIMFA
midt <- MIMFA(midt, ncomp=3, M=10)
midt

#-- estimates the number of MFA components for data imputation
#-- ncomp is chosen to being enough large
## Not run: 
midt <- MIMFA(midt, ncomp=50, M=10, estimeNC=TRUE)
midt
## End(Not run)

GonzalezIgnacio/HandlingMissRows documentation built on Jan. 17, 2020, 6:29 p.m.