reproducer (version 0.5.2)

calculate2GMdMRE: calculate2GMdMRE

Description

The function uses simulations assess the accuracy and power of parametric and non-parametric effect sizes for two group designs and four different distributions.

Usage

calculate2GMdMRE(
  mean = 0,
  sd = 1,
  N,
  reps,
  diff = c(0.2, 0.5, 0.8),
  type = "n",
  seed = 123,
  StdAdj = 0,
  AlwaysTwoSidedTests = FALSE,
  LargeSampleSize = 5e+06
)

Value

Tibble identifying the accuracy i.e. MdMRE of three effect sizes Cliff's d, CentralPHat, and StdMD estimates found in each sample, the variance of each of effect sizes, the mean of the observed values, the expected values of the effect sizes based on a single large sample and the power of each effect size based on the proportion of samples for which the effect sizes were significant.

Arguments

mean

The mean (or rate for gamma data) of the baseline distribution

sd

The standard deviation or shape of the baseline distribution

N

The nuber of obervations per group in each of the simulated experiment

reps

The number of experiments (i.e. two-group datasets) to be simulated

diff

A list identifying the mean group differences being assessed

type

A string parameter defining the distribution being simulated i.e. 'n' for normal data, 'l' for log-normal data, 'g' for gamma data and 'lap' for LaPlace data.

seed

A starting value for the simulations

StdAdj

A value that can be used to add additional variance for normal, lognormal and Laplce data and to change the shape parameter for gamma data.

AlwaysTwoSidedTests

A boolean variable. If TRUE the simulation always used two-sided tests otherwise the simulations use one-sided tests.

LargeSampleSize

Size of the large sample (default 5000000) used in each of the two groups used to obtain expected values of the non-parametric effect sizes (the expected value of the standardized men difference are always known)

Author

Barbara Kitchenham and Lech Madeyski

Examples

as.data.frame(
  calculate2GMdMRE(
    mean=0, sd=1, N=10, reps=20, diff=c(0.2,0.5,0.8), type='n', seed=123,
    StdAdj = 0, AlwaysTwoSidedTests=FALSE, LargeSampleSize=10000))
#   Design Obs Diff CliffdMdMRE  CentralPHatMdMRE StdESMdMRE  varCliffd
# 1   2G_n  20  0.2   1.8866607        -0.3191790 1.3757827 0.06089579
# 2   2G_n  20  0.5   0.5956595        -0.3735555 0.5295375 0.04872737
# 3   2G_n  20  0.8   0.3077882        -0.4086031 0.3698596 0.03492526
#       varPHat  StdESVar ObsCliffd ObsPHat  ObsStdES CliffdExpected PHatExpected
# 1 0.015223947 0.1666288     0.127  0.5635 0.2267731      0.1060074    0.5530037
# 2 0.012181842 0.1804443     0.283  0.6415 0.5386875      0.2694811    0.6347405
# 3 0.008731316 0.1978285     0.429  0.7145 0.8506020      0.4223684    0.7111842
# StdESExpected CliffdPower PHatPower StdESPower
# 1     0.1866728        0.10      0.10        0.1
# 2     0.4884736        0.15      0.15        0.2
# 3     0.7882317        0.50      0.50        0.6
#as.data.frame(
#  calculate2GMdMRE(
#   mean=0, sd=1, N=10, reps=100, diff=c(0.2,0.5,0.8), type='n', seed=123,
#   StdAdj = 0, AlwaysTwoSidedTests=FALSE, LargeSampleSize=10000))
#  Design Obs Diff CliffdMdMRE CentralPHatMdMRE StdESMdMRE  varCliffd
#1   2G_n  20  0.2   1.9702583       -0.3144569  1.7111237 0.07552663
#2   2G_n  20  0.5   0.6428482       -0.3641988  0.6357666 0.06842711
#3   2G_n  20  0.8   0.4036555       -0.3803787  0.4063333 0.05873046
#     varPHat  StdESVar ObsCliffd    ObsPHat  ObsStdES CliffdExpected
#1 0.01888166  0.2522504    0.1092    0.5546 0.1835910      0.1030653
#2 0.01710678  0.2624379    0.2646    0.6323 0.4949159      0.2678275
#3 0.01468262  0.2787054    0.4078    0.7039 0.8062407      0.4211512
#  PHatExpected StdESExpected CliffdPower PHatPower StdESPower
#1    0.5515327     0.1870328        0.09      0.12       0.10
#2    0.6339138     0.4869850        0.18      0.22       0.24
#3    0.7105756     0.7869372        0.44      0.52       0.56