constants: "Constants" Used in Program Control

Description Details Value

Description

Values used in controlling the executions of various functions in the lcmix package.

Details

All “constants” should be treated as though they were actual constants, and never reassigned. For example, strenuously avoid statements such as LC_EPS <- 1e-20.

Value

LC_LOG2

log(2), approximately 0.6931472.

LC_LOGPI

log(pi), approximately 1.14473.

LC_LOG2PI

log(2*pi), approximately 1.837877.

LC_LOGSQRT2PI

log(sqrt(2*pi)), approximately 0.9189385.

LC_LOGSQRT2

log(sqrt(2)), approximately 0.3465736.

LC_EPS

“epsilon”, a number very near 0. Defined here as sqrt(.Machine$double.eps), or approximately 1.490116e-08 on most systems.

LC_LEPS

the logarithm of LC_EPS.

LC_REPS

“reverse epsilon”, a number very near 1. Defined here as 1-LC_EPS.

LC_LREPS

the logarithm of LC_REPS.

LC_MAXSTDNORM (LC_MINSTDNORM)

maximum (minimum) values for numbers supposed to have been generated by a standard normal distribution, defined here as 8 (-8). Values greater (less) than this will have a standard normal density very near 0.

LC_ITER_MAX

default maximum iterations for the EM algorithm as used in mixmod and mdmixmod, defined here as 1000.

LC_ITER_TOL

the value below which relative changes in log-likelihood must fall to terminate the EM algorithm, defined here as 1e-6.

LC_TOPOLOGY

vector of allowable topology names for multiple-data mixture models fitted using mdmixmod, currently "layered" and "chained".

LC_FAMILY

a list of which each element bears the name of an allowable distribution family for mixture models. Each list element is itself a two-element list with elements uni and multi, corresponding to the names of the univariate and multivariate versions of the distribution family. The current families, and univariate and multivariate names, are given in the following table. The Pearson Type VII family is referred to as “pvii”.

family univariate multivariate
normal norm mvnorm
pvii pvii mvpvii
weibull weisd mvweisd
gamma gamma mvgamma
exponential exp mvexp
LC_DISTN

names of acceptable model distributions, from the “univariate” and “multivariate” columns of LC_FAMILY.

LC_NONNEGFAM

a vector of names of distribution families which require data to have non-negative values, currently "weibull", "gamma", and "exponential".

LC_SIMPAR

marginal distribution parameters for simulated mixture models. See simulation for examples of usage.


lcmix documentation built on May 2, 2019, 6:49 p.m.