Utility Functions in sirt — sirt-utilities • sirt

Utility functions in sirt.

Usage

# bounds entries in a vector
bounds_parameters( pars, lower=NULL, upper=NULL)

# improper density function which always returns a value of 1
dimproper(x)

# generalized inverse of a symmetric function
ginverse_sym(A, eps=1E-8)
# hard thresholding function
hard_thresholding(x, lambda)
# soft thresholding function
soft_thresholding(x, lambda)

# power function x^a, like in Cpp
pow(x, a)
# trace of a matrix
tracemat(A)

#** matrix functions
sirt_matrix2(x, nrow)   # matrix() function with byrow=TRUE
sirt_colMeans(x, na.rm=TRUE)
sirt_colSDs(x, na.rm=TRUE)
sirt_colMins(x, na.rm=TRUE)
sirt_colMaxs(x, na.rm=TRUE)
sirt_colMedians(x, na.rm=TRUE)

#* normalize vector to have sum of one
sirt_sum_norm(x, na.rm=TRUE)
#* discrete normal distribution
sirt_dnorm_discrete(x, mean=0, sd=1, ...)

# plyr::rbind.fill implementation in sirt
sirt_rbind_fill(x, y)

# Fisher-z transformation, see psych::fisherz
sirt_fisherz(rho)
# inverse Fisher-z transformation, see psych::fisherz2r
sirt_antifisherz(z)

# smooth approximation of the absolute value function
sirt_abs_smooth(x, deriv=0, eps=1e-4)

# permutations with replacement
sirt_permutations(r,v)
  #-> is equivalent to gtools::permutations(n=length(v), r=D, v=v, repeats.allowed=TRUE)

# attach all elements in a list in a specified environment
sirt_attach_list_elements(x, envir)

# switch between stats::optim and stats::nlminb
sirt_optimizer(optimizer, par, fn, grad=NULL, method="L-BFGS-B", hessian=TRUE,
                   control=list(), ...)

# print objects in a summary
sirt_summary_print_objects(obji, from=NULL, to=NULL, digits=3, rownames_null=TRUE,
      grep_string=NULL)
# print package version and R session
sirt_summary_print_package_rsession(pack)
# print package version
sirt_summary_print_package(pack)
# print R session
sirt_summary_print_rsession()
# print call
sirt_summary_print_call(CALL)

# print a data frame x with fixed numbers of digits after the decimal
print_digits(x, digits=NULL)

# discrete inverse function
sirt_rcpp_discrete_inverse(x0, y0, y)

# move variables in a data frame
move_variables_df(x, after_var, move_vars)

Arguments

pars: Numeric vector
lower: Numeric vector
upper: Numeric vector
x: Numeric vector or a matrix or a list
eps: Numerical. Shrinkage parameter of eigenvalue in ginverse_sym
a: Numeric vector
lambda: Numeric value
A: Matrix
nrow: Integer
na.rm: Logical
mean: Numeric
sd: Numeric
y: Matrix
rho: Numeric
deriv: Integer indicating the order of derivative
z: Numeric
r: Integer
v: Vector
envir: Environment
optimizer: Can be one of the following optimizers: optim, nlminb, bobyqa (from the minqa packlage), Rvmmin (from the optimx package) or nloptr (from the nloptr package using the argument opts$algorithm="NLOPT_LD_MMA").
par: Initial parameter
fn: Function
grad: Gradient function
method: Optimization method
hessian: Logical
control: Control list for R optimizers
...: Further arguments to be passed
obji: Data frame
from: Integer
to: Integer
digits: Integer
rownames_null: Logical
grep_string: String
pack: Package name
CALL: Call statement
x0: Vector
y0: Vector
after_var: String indicating variable name after which variable specified variables in move_vars should be moved
move_vars: Variables which should be moved after after_var

Examples

#############################################################################
## EXAMPLE 1: Trace of a matrix
#############################################################################

set.seed(86)
A <- matrix( stats::runif(4), 2,2 )
tracemat(A)
sum(diag(A))    #=sirt::tracemat(A)

#############################################################################
## EXAMPLE 2: Power function
#############################################################################

x <- 2.3
a <- 1.7
pow(x=x,a=a)
x^a            #=sirt::pow(x,a)

#############################################################################
## EXAMPLE 3: Soft and hard thresholding function (e.g. in LASSO estimation)
#############################################################################

x <- seq(-2, 2, length=100)
y <- sirt::soft_thresholding( x, lambda=.5)
graphics::plot( x, y, type="l")

z <- sirt::hard_thresholding( x, lambda=.5)
graphics::lines( x, z, lty=2, col=2)

#############################################################################
## EXAMPLE 4: Bounds on parameters
#############################################################################

pars <- c(.721, .346)
bounds_parameters( pars=pars, lower=c(-Inf, .5), upper=c(Inf,1) )

#############################################################################
## EXAMPLE 5: Smooth approximation of absolute value function
#############################################################################

x <- seq(-1,1,len=100)
graphics::plot(x, abs(x), lwd=2, col=1, lty=1, type="l", ylim=c(-1,1) )
# smooth approximation
tt <- 2
graphics::lines(x, sirt::sirt_abs_smooth(x), lty=tt, col=tt, lwd=2)
# first derivative
tt <- 3
graphics::lines(x, sirt::sirt_abs_smooth(x, deriv=1), lty=tt, col=tt, lwd=2)
# second derivative
tt <- 4
graphics::lines(x, sirt::sirt_abs_smooth(x, deriv=2), lty=tt, col=tt, lwd=2)

# analytic computation of first and second derivative
stats::deriv( ~ sqrt(x^2 + eps), namevec="x", hessian=TRUE )

if (FALSE) {
#############################################################################
## EXAMPLE 6: Permutations with replacement
#############################################################################

D <- 4
v <- 0:1
sirt::sirt_permutations(r=D, v=v)
gtools::permutations(n=length(v), r=D, v=v, repeats.allowed=TRUE)
}