replext_t4_c6.1 {npboottprm}

Replicate and Extend Simulation Results from Table 4 Cell 6.1

Description

This function, a specialized variant of 'replext_t4_c1.1', is designed to replicate and extend the simulation results from Table 4 cell 6.1 of the paper by Dwivedi et al. (2017). It employs different distributions for the two groups, using Chi-squared and Poisson distributions respectively, in line with the specific cell conditions.

Usage

replext_t4_c6.1(
  rdist = c("rchisq", "rpois"),
  par1_1 = 6,
  par2_1 = 0,
  par1_2 = 10,
  par2_2 = NULL,
  n1 = c(5, 5, 10),
  n2 = c(5, 10, 10),
  n_simulations = 10000,
  nboot = 1000,
  conf.level = 0.95
)

Arguments

rdist

Vector of distribution types, with the defaults set to 'rchisq' (Chi-squared) for the first group and 'rpois' (Poisson) for the second group. Other options include 'rlnorm' (lognormal) and 'rcauchy' (Cauchy).

par1_1

First parameter for the first group's distribution, default is 6 for Chi-squared's degrees of freedom.

par2_1

Second parameter for the first group's distribution, typically 0 for Chi-squared.

par1_2

First parameter for the second group's distribution, default is 10 for Poisson's lambda.

par2_2

Second parameter for the second group's distribution, typically NULL for Poisson.

n1

Vector of sample sizes for the first group.

n2

Vector of sample sizes for the second group, must be the same length as n1.

n_simulations

Number of simulations to run, default is 10,000.

nboot

Number of bootstrap samples, default is 1000.

conf.level

Confidence level for calculating p-value thresholds, default is 0.95.

Value

A data frame with columns for each sample size pair (n1, n2) and the proportions of significant p-values for each test (ST, WT, NPBTT, WRST, PTTa, PTTe).

Note

When using rlnorm (lognormal distribution), 'par1' represents 'meanlog' (the mean of the logarithms) and 'par2' represents 'sdlog' (the standard deviation of the logarithms). For rpois (Poisson distribution), 'par1' is 'lambda' (the rate parameter). In the case of rchisq (Chi-squared distribution), 'par1' is 'df' (degrees of freedom) and 'par2' is typically 0 as 'ncp' (non-centrality parameter) is not often used. Lastly, for rcauchy (Cauchy distribution), 'par1' is the 'location' parameter and 'par2' is the 'scale' parameter.

References

Dwivedi AK, Mallawaarachchi I, Alvarado LA. Analysis of small sample size studies using nonparametric bootstrap test with pooled resampling method. Stat Med. 2017 Jun 30;36(14):2187-2205. doi: 10.1002/sim.7263. Epub 2017 Mar 9. PMID: 28276584.

Examples

replext_t4_c6.1(n1 = c(10), n2 = c(10), n_simulations = 1)