steveharoz

library(tidyverse)
library(patchwork)
library(glue)

set.seed(7)

data = tibble(
  x = rnorm(50, 5),
  y = x + rnorm(50, 0, .7),
  x_noise = x + rnorm(50, 0, 1),

steveharoz

# t-test vs wilcox vs ordinal

library(tidyverse)
library(multidplyr) # parallelize
library(rms) # ordinal regression


sample_size = 100 # N

# genarate paired sets and calculate p-values with different techniques

steveharoz

library(tidyverse)
library(ggdist)



data = expand.grid(
  effect_size = c(0, 0.2, 0.4, 0.6, 0.8),
  sample_size = 1000,
  rep = 1:10000
) %>% 

steveharoz

# put each attribute into a column if possible
# rownames_to_column: (logical) add rownames if the attribute exists
# fix_name_conflicts: (logical) if an attribute name already exists as a column name, find a unique name for it
attributes_to_columns = function(dataframe, rownames_to_column = TRUE, fix_name_conflicts = TRUE) {
    for (name in names(attributes(dataframe))) {
      
      value = attr(dataframe, name)
      
      if (name %in% c("names", "class")) {
        #skip

steveharoz

Original idea

BlueSky thread

steveharoz

library(tidyverse)

# simulate one experiment
simulate = function(subject_count = 20, effect_size = 0) {
  a = rnorm(subject_count/2)
  b = rnorm(subject_count/2, effect_size)
  tibble(
    p = t.test(a, b, alternative = "less")$p.value,
    d = (mean(b) - mean(a)) / sd(c(a-mean(a), b-mean(b)))
  )

steveharoz

# t-test vs wilcox vs ordinal

library(tidyverse)
library(multidplyr) # parallelize
library(rms) # ordinal regression


sample_size = 500 # N

# generate paired sets and calculate p-values with different techniques

steveharoz

steveharoz

# Simulate different experiment designs and approaches to calculating Cohen's D

# References
# https://journalofcognition.org/articles/10.5334/joc.10   (h/t Aaron Caldwell @arcstats.bsky.social)
# https://jakewestfall.org/blog/index.php/2016/03/25/five-different-cohens-d-statistics-for-within-subject-designs/


library(tidyverse)
library(lmerTest)
library(multidplyr) # not necessary, but helps performance

steveharoz

library(tidyverse)
library(effectsize) # cohens_d()

# reproducible
set.seed(8675309)

# simulate one experiment
simulate = function(count_a, count_b, replicate_count) {
  
  count_total = count_a + count_b