setOptions: Low-level options of a SIMoNe run

Description

This function is intended to design low-level uses of SIMoNe by specifying various parameters of the underlying algorithms.

Usage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
setOptions(normalize      = TRUE,
           verbose        = TRUE,
           penalties      = NULL,
           penalty.min    = NULL,
           penalty.max    = NULL,
           n.penalties    = 100, 
           edges.max      = Inf,
           edges.sym.rule = NULL,
           edges.steady   = "neighborhood.selection",
           edges.coupling = "coopLasso",
           clusters.crit  = "BIC",
           clusters.meth  = "bayesian",
           clusters.qmin  = 2,
           clusters.qmax  = 4)

Arguments

normalize

logical specifying wether the data should be normalized to unit variance. The normalization is made task-wisely in the multiple sample setting. Default is TRUE.

verbose

a logical that indicates verbose mode to display progression. Default is TRUE.

penalties

vector of decreasing penalty levels for the network estimation. If NULL (the default), an appropriate vector will be generated in simone with n.penalties entries, starting from penalty.max and shrinked to penalty.min.

penalty.min

The minimal value of the penalty that will be tried for network inference. If NULL (the default), it will be set in simone to 1e-5 for the monotask framework and to 1e-2 for the multitask framework.

penalty.max

The maximal value of the penalty that will be tried for network inference. If NULL (the default), it will be set to a value that provoques an empty granph. Default is NULL.

n.penalties

integer that indicates the number of penalties to put in the penalties vector. Default is 100.

edges.max

integer giving an upper bound for the number of edges to select: if a network is inferred along the algorithm with a number of edges overstepping edges.max, it will stop there. Default is Inf.

edges.steady

a character string indicating the method to use for the network inference associated to steady-state data, one task framework. Either "graphical.lasso" or "neighborhood.selection". Default is the later.

edges.coupling

character string (either "coopLasso", "groupLasso" or "intertwined") that indicates the coupling method across task in the multiple sample setup. Defautl is "coopLasso".

edges.sym.rule

character string ("AND", "OR", "NO") for post-symmetrization of the infered networks. Enforced to "NO" for time-course data (directed network) and set to "AND" as default for steady-state data (undirected network).

clusters.crit

criterion to select the network that is used to find an underlying clustering. Either "BIC", "AIC" or an integer for the number of edges. Default is "BIC".

clusters.qmin

minimum number of classes for clustering. Default is 2.

clusters.qmax

maximum number of classes for clustering. Default is 4.

clusters.meth

character string indicating the strategy used for the estimation: "variational", "classification", or "bayesian". See the mixer package for further details. Default is "bayesian".

Value

A list that contains all the specified parameters.

Note

If the user specifies its own penalties vector, all the networks inferred during the algorithm will be kept, even if they share the very same number of edges.

On the other hand, if you only specify penalty.max and/or penalty.min and/or n.penalties, the algorithm will only kept the networks who show different numbers of edges. That is to say, the number of networks stocked in the output of simone generally does not have a length equal to n.penalties.

Author(s)

J. Chiquet

See Also

simone.

Examples

1
2
## generate an object (list) with the default parameters
setOptions()

Questions? Problems? Suggestions? or email at ian@mutexlabs.com.

All documentation is copyright its authors; we didn't write any of that.