experimentD: Running a computer experiment

In metagen: Inference in Meta Analysis and Meta Regression

Description

Runs a computer experiment that evaluates the performance of different inference methods for the random effects meta regression model with respect to heterogeneity and regression coefficients.

Usage

  experimentD(n, h, d, s, x, b, sgnf, piv_draws)

Arguments

n	number of draws.
h	heterogeneity.
d	heteroscedasticity.
s	vector study sizes.
x	design matrix.
b	regression coefficients.
sgnf	significance levels.
piv_draws	privotal draws.

Details

This also includes methods adjusting for uncertainty in the heteroscedasticity vector. In particular, the study sizes need to be known, here.

Value

Data frame of accumulated performance measures.

Examples

h_test <- 0.03
x_test <- cbind(1,1:7)
b_test <- c(.5, .25)
sgnf_test <- c(0.025, 0.01)

set.seed(5133568) # for reproducibility
d_test <- rchisq(7, df=0.02)
s_test <- runif(7, min=200, max=2000)

# In an actual computer experiment, use 'piv_draws=1000' instead!!
experimentD(n=5, h=h_test, d=d_test, s=s_test, x=x_test, b=b_test,
  sgnf=sgnf_test, piv_draws=50)

metagen documentation built on May 2, 2019, 6:08 a.m.