ocoptions: Options for the opticut Package

In psolymos/opticut: Likelihood Based Optimal Partitioning and Indicator Species Analysis

Options for the opticut Package

Description

A convenient way of handling options related to the opticut package.

Usage

ocoptions(...)

Arguments

...	arguments in tag = value form, or a list of tagged values. The tags must come from the parameters described below.

Value

When parameters are set by ocoptions, their former values are returned in an invisible named list. Such a list can be passed as an argument to ocoptions to restore the parameter values. Tags are the following:

collapse	character value to be used when merging factor levels, the default is "+".
cut	log likelihood ratio value, model/species with lower values are excluded from summaries and plots, the default is 2.
sort	logical value indicating if species/partitions should be meaningfully sorted, the default is TRUE. It can take numeric value when only species (1) or partitions (2) are to be sorted (1:2 is equivalent to TRUE).
theme	the color theme to be used based on occolors, the default is "br".
check_comb	check the design matrices for complementary partitions using checkComb, the default is TRUE.
try_error	if opticut and multicut should try to exclude species where the models failed (TRUE), the default is to stop when an error is encountered (FALSE).
scale	the scaling factor used to calculate indicator potential (I) based on the estimated contrast (x): I = abs(tanh(x*scale)), the default is 0.5.
fix_fitted	bestpart.multicut uses lorenz which requires nonnegative fitted values, however models with identity link can lead to negative expected values. When TRUE the fitted values (x) are adjusted as x' = x + abs(min(x)) to ensure nonnegativity. The default is FALSE.
robust_loglik	if ill-defined models resulting in perfect fit (infinite log likelihood, or NA, NaN) should be allowed. The default TRUE makes such ill-defined log likelihoods a very small real number -(.Machine$double.xmax^(1/3)). FALSE is equivalent to allowing every model to safeguard against such cases or not.

Author(s)

Peter Solymos <[e-mail]>

Examples

## simple example from Legendre 2013
## Indicator Species: Computation, in
## Encyclopedia of Biodiversity, Volume 4
## http://dx.doi.org/10.1016/B978-0-12-384719-5.00430-5
gr <- as.factor(paste0("X", rep(1:5, each=5)))
spp <- cbind(Species1=rep(c(4,6,5,3,2), each=5),
    Species2=c(rep(c(8,4,6), each=5), 4,4,2, rep(0,7)),
    Species3=rep(c(18,2,0,0,0), each=5))
rownames(spp) <- gr
## must add some noise to avoid perfect fit
spp[6, "Species1"] <- 7
spp[1, "Species3"] <- 17
spp

## current settings
print(unlist(ocoptions())) # these give identical answers
unlist(getOption("ocoptions"))
summary(ocall <- opticut(spp ~ 1, strata=gr, dist="gaussian", comb="all"))

## resetting pboptions and checking new settings
ocop <- ocoptions(collapse="&", sort=FALSE)
unlist(getOption("ocoptions"))
## running again with new settings
summary(ocall <- opticut(spp ~ 1, strata=gr, dist="gaussian", comb="all"))

## resetting original
ocoptions(ocop)
unlist(getOption("ocoptions"))

psolymos/opticut documentation built on Nov. 27, 2022, 11:29 a.m.