gistfile1.txt

library(broom)
library(purrr)
library(dplyr)
library(ggplot2)

run_spurious_regression <- function(nvars = 6, nobs = 1000){
  dataset <- as.data.frame(replicate(nvars, rnorm(nobs))) 
  tidy(lm(V1 ~ ., data = dataset))
}

many_spurious_regressions <- rerun(1, run_spurious_regression()) %>%
  map2(.x = ., .y = seq(1, 1), ~mutate(.x, run = paste0("Run number ", .y))) %>%
  bind_rows()

many_spurious_regressions %>%
  mutate(is_significant = ifelse(p.value < .05, "Yes", "No")) %>%  
  ggplot() +
  geom_point(aes(y = estimate, x = term, colour = is_significant)) +
  labs(title = "Significative results in totally randomly generated data",
       subtitle = "6 normally distributed variables of size 1000 were created, and the first was regressed on the 5 others.
The plots show the coefficients from one particularly favorable regression. The colour of the dots show if the coefficients were significant at the 5% level.") +
  theme_minimal() +
  theme(legend.position = "bottom")



many_spurious_regressions <- rerun(100, run_spurious_regression()) %>%
  map2(.x = ., .y = seq(1, 100), ~mutate(.x, run = paste0("Run number ", .y))) %>%
  bind_rows()

many_spurious_regressions %>%
  mutate(is_significant = ifelse(p.value < .05, "Yes", "No")) %>%  
  ggplot() +
  geom_point(aes(y = estimate, x = term, colour = is_significant)) +
  labs(title = "Significative results in totally randomly generated data",
       subtitle = "6 normally distributed variables of size 1000 were created, and the first was regressed on the 5 others.
The plots show the coefficients from 100 runs and their colours show if these coefficients were significant at the 5% level.") +
  theme_minimal() +
  theme(legend.position = "bottom")