rubenarslan

## dumbed-down simulation of selection bias with noise (aka selecting on fielding of scale before fielding)

# set seed for reproducibility
set.seed(123)

# parameters
n_studies <- 100000
noise_dist <- seq(0.01, 0.07, 0.01)

## draw distribution of true alpha values

rubenarslan

library(tidyverse)
# simulate unobserved trait and items with error
N <- 100
dat <- tibble(
  id = 1:N,
  trait = rnorm(N),
  item1 =  trait + rnorm(N),
  item2 =  trait + rnorm(N),
  item3 =  trait + rnorm(N),
  item4 =  trait + rnorm(N),

rubenarslan

library(dplyr)
N = 10000
people <- tibble(
  male = sample(0:1, N, TRUE),
  trait = rnorm(N) + male)
true_means <- people %>% 
  group_by(male) %>% 
  summarise(sex_mean = mean(trait))
people <- people %>% left_join(true_means) %>% 
  left_join(true_means %>% rename(opposite_sex_mean = sex_mean) %>% mutate(male = 1-male)) %>% 

rubenarslan

N <- 5000
sim <- tibble(
  group =  rep(0:1, each = N/2),
  evb_strength =  0.1 + group * 0.9,
  # four orthogonal traits
  trait1 = rnorm(N),
  trait2 = rnorm(N),
  trait3 = rnorm(N),
  trait4 = rnorm(N),
  # evaluative bias

rubenarslan

options(width=200)
library(tidyverse)
gpt <- rio::import("~/Downloads/tmp3.csv")

items <- tibble(var=c(str_c("ipipc", 1:10),  str_c("grit", 1:10)), 
                label = names(gpt)[-1])
knitr::kable(items)

rubenarslan

n <- 1000
people <- tibble(
	children = rpois(n, 1.4),
	# for childless people, sat with childcare is not applicable
	sat_with_child_care = if_else(children > 0, rnorm(n), NA_real_),
	sat_with_housing = rnorm(n),
	happiness = rnorm(n) + sat_with_housing + 
		if_else(children > 0, sat_with_child_care, 0),
)

rubenarslan

mean_se_cluster <- function (x, mult = 1, cluster = NULL) 
{
    x_na <- is.na(x)
    x <- x[!x_na]
    cluster <- cluster[!x_na]
    stopifnot(!is.null(cluster))
    mod <- lme4::lmer(x ~ 1 + (1 | cluster))
    intercept <- broom.mixed::tidy(mod, effects = "fixed")
    se <- mult * intercept$std.error
    mean <- intercept$estimate

rubenarslan

library(tidyverse)

n_papers <- 1000
papers <- tibble(
  paper = 1:n_papers,
  quality = rnorm(n_papers),
  reviews = 0,
  published = FALSE,
  journal = NA_real_,
  most_recent_assessment = NA_real_

rubenarslan

library(brms)
library(dplyr)
n = 1000
fakedata = data_frame(
  # structure
  iv = rnorm(n),
  dv = 0.5 * iv + 0.5 * rnorm(n),
  # measurement
  iv_m = iv + 1 * rnorm(n),
  dv_m = dv + 1 * rnorm(n)

rubenarslan

# Manual page
library(curl)
?multi

# Example of 100 requests over 10 connections
real = vector("list", 10)
for(i in 1:10){
    h <- new_handle(url = "https://httpbin.org/get")
    multi_add(h, complete = function(res){