mpower (version 0.1.0)

sim_curve: Power curve using Monte Carlo simulation

Description

This function can be used to create power curves by calling sim_power() on combinations of many sample sizes and signal-to-noise ratio (SNR).

Usage

sim_curve(
  xmod,
  ymod,
  imod,
  s = 100,
  n = 100,
  cores = 1,
  file = NULL,
  errorhandling = "stop",
  snr_iter = 10000,
  cluster_export = c()
)

Value

A SimCurve object with the following attributes:

s

a number of simulations.

snr

a real number or array of real numbers for SNR of each OutcomeModel.

n

a number or vector of sample sizes.

xmod

the MixtureModel used.

ymod

the OutcomeModel used.

imod

the InferenceModel used.

sims

a list of simulation output matrices.

Arguments

xmod

A MixtureModel object.

ymod

One or a list of OutcomeModel object(s).

imod

An InferenceModel object.

s

An integer for the number of Monte Carlo simulations.

n

An integer or a vector of sample sizes.

cores

An integer for the number of processing cores. When cores > 1, parallelism is automatically applied.

file

A string, a file name with no extension to write samples to periodically. By default write to an RDS file.

errorhandling

A string "remove", "stop", or "pass". If an error occurs in any iteration, remove that iteration (remove), return the error message verbatim in the output (pass), or terminate the loop (stop). Default is "remove". See R package 'foreach' for more details.

snr_iter

An integer for number of Monte Carlo samples to estimate SNR.

cluster_export

A vector of functions to pass to the parallel-processing clusters.

Examples

data("nhanes1518")
chems <- c("URXCNP", "URXCOP", "URXECP", "URXHIBP", "URXMBP", "URXMC1",
"URXMCOH", "URXMEP","URXMHBP", "URXMHH", "URXMHNC", "URXMHP", "URXMIB",
"URXMNP", "URXMOH", "URXMZP")
chems_mod <- mpower::MixtureModel(nhanes1518[, chems], method = "resampling")
bmi_mod <- mpower::OutcomeModel(f = "0.2*URXCNP + 0.15*URXECP +
0.1*URXCOP*URXECP", family = "binomial")
logit_mod <- mpower::InferenceModel(model = "glm", family = "binomial")
logit_out <- mpower::sim_curve(xmod=chems_mod, ymod=bmi_mod, imod=logit_mod,
s=20, n=c(500, 1000), cores=2, snr_iter=1000)
logit_df <- summary(logit_out, crit="pval", thres=0.05, how="lesser")
plot_summary(logit_out, crit="pval", thres=0.05, how="lesser")