TJ Mahr tjmahr

d <- posterior::example_draws()
library(posterior)
#> Warning: package 'posterior' was built under R version 4.6.1
#> This is posterior version 1.7.1
#> 
#> Attaching package: 'posterior'
#> The following objects are masked from 'package:stats':
#> 
#>     mad, sd, var

t <- tempfile(fileext = ".R")
writeLines(con = t, text = r"(
#' Get a function's source code and any preceding comment lines
#' @param fun a function with a `srcref` attribute
#' @return a character vector of source code for the function
get_function_source_with_comment_block <- function(fun) {
 stopifnot(!is.null(attr(fun, "srcref")))

 source_lines <- fun |>

withr::local_temp_libpaths()
pak::pkg_install("ggplot2@3.5.2")
#> ℹ Loading metadata database✔ Loading metadata database ... done
#>  
#> → Will install 20 packages.
#> → Will download 18 CRAN packages (11.47 MB), cached: 2 (1.51 MB).
#> + cli            3.6.5  [dl] (1.40 MB)
#> + farver         2.1.2  [dl] (1.52 MB)
#> + ggplot2        3.5.2  [bld]

library(tidyverse)

TLDR: probability functions like pt() can provide log values so that you can avoid numerical problems. (This is why we sum log likelihoods instead of multiplying likelihoods.) So, we can identify the FLOATING-POINT BREAKERS of p-values if we stick to the log scale throughout the pipeline.

library(tidyverse)
# d <- readr::read_tsv("test.tsv") |>
#   janitor::clean_names()
#
# d$total <- d$biden + d$trump + d$other
# d$prop_votes_in <- d$ballots_accepted / d$total
# d$prop_biden <- d$biden / d$total
# datapasta::tribble_paste(d)

library(ggplot2)

geom_crossrange <- function(
    mapping = NULL,
    data = NULL,
    stat = "identity",
    position = "identity",
    ...,
    na.rm = FALSE,

print.file_check_log <- function(x, ...) str(x, ...)
as_file_check_log <- function(x) UseMethod("as_file_check_log")
as_file_check_log.default <- function(x) {
  structure(list(x = x, notes = list()), class = "file_check_log")
}

apply_to_file_check_log <- function(x, fn, ...) {
  results <- fn(x$x, ...)
  l <- list(results)

# Make a csv to read in
db1 <- DBI::dbConnect(duckdb::duckdb())
f <- tempfile("mtcars", fileext = ".csv")
write.csv(mtcars, f)

library(tidyverse)

# Read in the csv.
# Here two column types are hard-coded using a struct

Some of the rows have more columns than the first row of column names.

writeLines(
"a\tb\tc\td\te\tf
1\t2\t3\t4\t5\t6\t
1\t2\t3\t4\t5\t6\t7\t
1\t2\t3\t4\t5\t6\t7\t8
1\t2\t3\t4\t5\t6\t7\t8\t9\t10\t11
1\t2\t3\t4\t5\t6

	library(tidyverse)
	library(brms)
	targets::tar_load(model_wtocs_fast_phase_type_interaction_slopes)
	model <- model_wtocs_fast_phase_type_interaction_slopes
	data_marg <- compute_marginal_means_child_listener_slopes(model)


	# Compute marginal means for the model with random effects:
	# ~ (control_file_type \| subject_num) + (phase \| listener)
	compute_marginal_means_child_listener_slopes <- function(