ggm.estimate.pcor {GeneNet}

Graphical Gaussian Models: Small Sample Estimation of Partial Correlation

Description

ggm.estimate.pcor offers an interface to two related shrinkage estimators of partial correlation. Both are fast, statistically efficient, and can be used for analyzing small sample data.

The default method "statics" employs the function pcor.shrink whereas the "dynamic" method relies on dyn.pcor (in the longitudinal package). The difference between the two estimators is that the latter takes the spacings between time points into account if the input are multiple time course data (these must be provided as longitudinal object).

Usage

ggm.estimate.pcor(x, method = c("static", "dynamic"), ...)

Arguments

x

data matrix (each rows corresponds to one multivariate observation)

method

method used to estimate the partial correlation matrix. Available options are "static" (the default) and "dynamic" - both are shrinkage methods.

...

options passed to pcor.shrink and to dyn.pcor (in the longitudinal package).

Details

For details of the shrinkage estimators we refer to Opgen-Rhein and Strimmer (2006a,b) and Sch\"afer and Strimmer (2005), as well as to the manual pages of pcor.shrink (in the corpcor package) and dyn.pcor (in the longitudinal package).

Previously, this function offered several furthers options. The old option called "shrinkage" corresponds to the present "static" option. The other old options "observed.pcor", "partial.bagged.cor", and "bagged.pcor" are now considered obselete and have been removed.

Value

An estimated partial correlation matrix.

Author(s)

Rainer Opgen-Rhein, Juliane Sch\"afer, and Korbinian Strimmer (https://strimmerlab.github.io).

References

Opgen-Rhein, R., and K. Strimmer. 2006a. Inferring gene dependency networks from genomic longitudinal data: a functional data approach. REVSTAT 4:53-65.

Opgen-Rhein, R., and K. Strimmer. 2006b. Using regularized dynamic correlation to infer gene dependency networks from time-series microarray data. The 4th International Workshop on Computational Systems Biology, WCSB 2006 (June 12-13, 2006, Tampere, Finland).

Sch\"afer, J., and Strimmer, K. (2005). A shrinkage approach to large-scale covariance estimation and implications for functional genomics. Statist. Appl. Genet. Mol. Biol. 4:32. <DOI:10.2202/1544-6115.1175>

See Also

ggm.simulate.data, ggm.estimate.pcor, pcor.shrink, and dyn.pcor (in the longitudinal package)

Examples

## Not run: 

# load GeneNet library
library("GeneNet")

# generate random network with 40 nodes 
# it contains 780=40*39/2 edges of which 5 percent (=39) are non-zero
true.pcor <- ggm.simulate.pcor(40)
  
# simulate data set with 40 observations
m.sim <- ggm.simulate.data(40, true.pcor)

# simple estimate of partial correlations
estimated.pcor <- cor2pcor( cor(m.sim) )

# comparison of estimated and true values
sum((true.pcor-estimated.pcor)^2)

# a slightly better estimate ...
estimated.pcor.2 <- ggm.estimate.pcor(m.sim)
sum((true.pcor-estimated.pcor.2)^2)


## End(Not run)