In animalSymbolicum/assumptions: Look up, plot, simulate and test statistical assumptions

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

The purpose of the 'assumptions' R-package is threefold:

• It tries to raise awareness of statistical assumptions. Before conducting a statistical test, its assumptions are printed to the console and the user is asked for confirmation after a graphical assessment.

• It collects tests and graphs to assess statistical assumptions and simplifies looking up specific assumptions.

• All the tests are made available for simulations including complex decision strategies for picking a statistical test for different diagnosed assumption violations.

Install package

To install 'assumptions' R-package from github run:

require(remotes)
remotes::install_github("animalSymbolicum/assumptions")
library(assumptions)

library(assumptions)
set.seed(33)

Hypothesis Tests

The assumptions package seeks to expand common statistical methods with the notice of their assumptions for educational purpose. To conduct a parametric or non-parametric hypothesis test call:

# t-test
asm_ttest(rnorm(20), color = F)

# wilcoxon test
asm_wilcox(rnorm(20), color = F)

Look Up assumptions

Common statistical methods assumptions can be looked up with:

# get assumptions of t.test
assumptions(t.test, color = F)

# get assumptions of wilcox.test
assumptions(wilcox.test, color = F)

Overview of assumptions assessment tests (pre-tests)

All tests and graphs to assess specific assumptions such as the distribution, independence, identical distribution (homogeneity) or randomness are collected in a central catalogue table.

asm_library()

Single assessment of an assumption

Single or combinations of assumptions can be assessed by calls like:

# only first test selected as example
asm_distribution(rnorm(10), 1)
asm_independence(rnorm(10), 1)
asm_randomness(rnorm(10), 1)
asm_homogeneity(list(A=rnorm(10), B=rnorm(10)), 1)

# a combination of tests
asm_preTests(rnorm(10), c("Shapiro-Wilk", "Rank", "Ljung-Pierce"))

Simulations

These tests are also available for simulations. Simulation can be defined and summarized by:

# minimal example
minSim <- asm_simulate(100, pre_selection = c("Shapiro-Wilk", "Rank", "Ljung-Pierce"))
asm_reportSim(minSim, report = "result")

Complex decision strategies can be defined in the simulation call or afterwards on simulation result:

# minimal example
asm_simStrategy(minSim, 
  list(
    simple  = quote(ifelse(pre_W_X & pre_W_Y, post_ttest, post_wilcox)),
    complex = quote(ifelse(sim_n_X < 11 & pre_W_X_pvalue <= 0.001 & !descr_skew_X > 1, post_ttest, post_wilcox))
  )
)
asm_reportSim(minSim, report = "result")

A simulation study like 'The two-sample t test: pre-testing its assumptions does not pay off' could be replicated with:

require(miceadds)
asm_simulate(
    simulations = 100000,
    sim_n = list(c(10,30,10,30,30), c(10,30,30,10,100)),
    sim_func = rep("asm_simData", 24),
    sim_args = mapply(function(distr, sig, delta) {
          list(
              distr = c("fleishman_sim",  "fleishman_sim"),
              dots  = list(
                  list(mean = delta, sd = sig, skew = distr[1], kurt = distr[2]),
                  list(skew = distr[3], kurt = distr[4])
              )
          )
      },
      rep(
        list(c(0,0,0,0), c(0,15,0,15), c(0.5,15,0.5,15), c(1,15,1,15), c(3,15,3,15), c(0,0,3,15)),
        4
      ),
      as.list(rep(rep(1:2, each = 6), 2)),
      as.list(rep(c(0, 1), each = 12)),
      SIMPLIFY=FALSE
    )
)

animalSymbolicum/assumptions documentation built on Dec. 19, 2021, 3:37 a.m.