p_glm: p-value from (generalized) linear regression model...

In Spower: Power Analyses using Monte Carlo Simulations

p-value from (generalized) linear regression model simulations with fixed predictors

Description

p-values associated with (generalized) linear regression model. Requires a pre-specified design matrix (X).

Usage

p_glm(
  formula,
  X,
  betas,
  test,
  sigma = NULL,
  family = gaussian(),
  gen_fun = gen_glm,
  ...
)

gen_glm(formula, X, betas, sigma = NULL, family = gaussian(), ...)

Arguments

formula	formula passed to either lm or glm
X	a data.frame containing the covariates
betas	vector of slope coefficients that match the model.matrix version of X
test	character vector specifying the test to pass to lht. Can also be a list of character vectors to evaluate multiple tests
sigma	residual standard deviation for linear model. Only used when family = 'gaussian'
family	family of distributions to use (see family)
gen_fun	function used to generate the required discrete data. Object returned must be a data.frame. Default uses gen_glm. User defined version of this function must include the argument ...
...	additional arguments to be passed to gen_fun. Not used unless a customized gen_fun is defined

Value

a single p-value

Author(s)

Phil Chalmers [e-mail]

See Also

p_lm.R2

Examples

X <- data.frame(G = factor(rep(c('control', 'treatment'), each=50)),
                C = sample(50:100, 100, replace=TRUE))
head(X)

# ANCOVA setup
p_glm(y ~ G + C, test="Gtreatment = 0",
  X=X, betas=c(10, .3, 1), sigma=1)

# ANCOVA setup with logistic regression
p_glm(y ~ G + C, test="Gtreatment = 0",
  X=X, betas=c(-2, .5, .01), family=binomial())

# ANCOVA setup with poisson regression
p_glm(y ~ G + C, test="Gtreatment = 0",
  X=X, betas=c(-2, .5, .01), family=poisson())



# test whether two slopes differ given different samples.
#   To do this setup data as an MLR where a binary variable S
#   is used to reflect the second sample, and the interaction
#   effect evaluates the magnitude of the slope difference
gen_twogroup <- function(n, dbeta, sdx1, sdx2, sigma, n2_n1 = 1, ...){
  X1 <- rnorm(n, sd=sdx1)
  X2 <- rnorm(n*n2_n1, sd=sdx2)
  X <- c(X1, X2)
  N <- length(X)
  S <- c(rep(0, n), rep(1, N-n))
  y <- dbeta * X*S + rnorm(N, sd=sigma)
  dat <- data.frame(y, X, S)
  dat
}

# prospective power using test that interaction effect is equal to 0
p_glm(formula=y~X*S, test="X:S = 0",
	  n=100, sdx1=1, sdx2=2, dbeta=0.2,
	  sigma=0.5, gen_fun=gen_twogroup) |> Spower(replications=1000)

Spower documentation built on June 8, 2025, 11:18 a.m.