Infusion (version 2.1.0)

options: Infusion options settings

Description

Allow the user to set and examine a variety of options which affect operations of the Infusion package. However, typically these should not be modified, and if they are, not more than once in a data analysis.

Usage

Infusion.options(...)
Infusion.getOption(x)

Value

For Infusion.getOption, the current value set for option x, or

NULL if the option is unset.

For Infusion.options(), a list of all set options. For

Infusion.options(name), a list of length one containing the set value, or NULL if it is unset. For uses setting one or more options, a list with the previous values of the options changed (returned invisibly).

Arguments

x

a character string holding an option name.

...

A named value or a list of named values. The following values, with their defaults, are used in Infusion:

mixturing

character string: package or function to be used for mixture modelling. Recognized packages are "Rmixmod" (the default) and "mclust";

train_cP_size:

Expression for train_cP_size argument of project.character.

trainingsize:

Expression for trainingsize argument of project.character.

projKnotNbr = 1000:

default value of trainingsize argument of project.character for REML (as implied by default expression for trainingsize).

logLname = "logL":

default value of logLname argument of infer_logLs. The name given to the inferred log likelihoods in all analyses.

LRthreshold= - qchisq(0.999,df=1)/2:

A value used internally by sample_volume to sample points in the upper region of the likelihood surface, as defined by the given likelihood ratio threshold.

precision = 0.1:

default value of precision argument of refine. Targets RMSE of log L and log LR estimates.

nRealizations=1000:

default value of nRealizations argument of add_simulation. Number of realizations for each empirical distribution.

mixmodGaussianModel="Gaussian_pk_Lk_Dk_A_Dk":

default models used in clustering by Rmixmod. Run Rmixmod::mixmodGaussianModel() for a list of possible models, and see the statistical documentation (Mixmod Team 2016) for explanations about them.

seq_nbCluster= function(projdata, nr=nrow(projdata)) {seq(ceiling(nr^0.3))}:

function to control the value of nbCluster used in clustering by Rmixmod (see Details for discussion of this default).

maxnbCluster = function(projdata) {...} :

function to control the maximum number of clusters (see Details).

example_maxtime=2.5:

Used in the documentation to control whether the longer examples should be run. The approximate running time of given examples (or some very rough approximation for it) on one author's laptop is compared to this value.

nb_cores

Number of cores for parallel computations (see Details for implementation of these).

gof_nstats_fn

See goftest.

and possibly other undocumented values for development purposes.

Details

The set of the number of clusters tried (nbCluster argument in Rmixmod) is controlled by two options: seq_nbCluster and maxnbCluster. The second is used to correct the first, using the dimensions of the projdata locally used for clustering, which typically differs from the dimensions of the user-level data (if projections have been applied, in particular). The default upper value of the nbCluster range is the value recommended in the mixmod statistical documentation (Mixmod Team, 2016). But this default may be suitable only for low-dimensional data, hence the need for correcting it bymaxnbCluster.

Infusion can perform parallel computations if several cores are available and requested though Infusion.options(nb_cores=.). If the doSNOW back-end is attached (by explicit request from the user), it will be used; otherwise, pbapply will be used. Both provide progress bars, but doSNOW may provide more efficient load-balancing. The character shown in the progress bar is 'P' for parallel via doSNOW backend, 'p' for parallel via pbapply functions, and 's' for serial via pbapply functions. In addition, add_simulation can parallelise at two levels: at an outer level over parameter point, or at an inner level over simulation replicates for each parameter point. The progress bar of the outer computation is shown, but the character shown in the progress bar is 'N' if the inner computation is parallel via the doSNOW backend, and 'n' if it is parallel via pbapply functions. So, one should see either 'P' or 'N' when using doSNOW.

References

Mixmod Team (2016). Mixmod Statistical Documentation. Université de Franche-Comté, Besançon, France. Version: February 10, 2016 retrieved from https://www.mixmod.org.

Examples

  Infusion.options()
  Infusion.getOption("LRthreshold")
  if (FALSE) {
  Infusion.options(LRthreshold=- qchisq(0.99,df=1)/2)
  }