The history of publishing delays by Daniel Himmelstein¶

library(dplyr, warn=F)
library(ggplot2)

journal_df %>%
  readr::write_tsv(file.path('data', 'journal-summaries.tsv'))

xlab_accept = 'Date accepted'
xlab_publish = 'Date published online'
ylab_accept = 'Acceptance delay (days)'
ylab_publish = 'Publication delay (days)'

plot_years <- function(percentile_df, delay_df, min_year, max_delay, xlab, ylab) {
  pctl_col = '#00693E'
  gg = percentile_df %>%
    ggplot2::ggplot(aes(x = year, y = delay, group = percentile)) +
    ggplot2::geom_smooth(aes(group=NULL), data=delay_df, color='#B4B3B3', fill = '#B4B3B3', alpha=1) +
    ggplot2::geom_line(size=0.1, color = pctl_col) +
    ggplot2::geom_line(data = dplyr::filter(percentile_df, percentile == 50), size=0.8, color = pctl_col) +
    ggplot2::geom_line(data = dplyr::filter(percentile_df, percentile %in% c(25, 75)), size=0.5, color = pctl_col) +
    ggplot2::coord_cartesian(xlim = c(min_year, 2015), ylim = c(0, max_delay), expand = FALSE) +
    ggplot2::theme_bw() +
    ggplot2::xlab(xlab) +
    ggplot2::ylab(ylab) +
    ggplot2::theme(plot.margin=grid::unit(c(2, 10, 2, 2), 'points'))
  return(gg)
}

# Acceptance delay plot
gg = plot_years(
  percentile_df = year_percentile_df %>% dplyr::filter(delay_type == 'Acceptance'),
  delay_df = delay_df %>% dplyr::filter(delay_type == 'Acceptance'),
  min_year = 1981,
  max_delay = 205,
  xlab = xlab_accept,
  ylab = ylab_accept
)

file.path('viz', 'acceptance-by-article.png') %>%
  ggplot2::ggsave(plot = gg, width = 5.5, height = 0.75 * 5.5)
file.path('viz', 'acceptance-by-article.pdf') %>%
  ggplot2::ggsave(plot = gg, width = 5.5, height = 0.75 * 5.5)

# Publication delay plot
gg = plot_years(
  percentile_df = year_percentile_df %>% dplyr::filter(delay_type == 'Publication'),
  delay_df = delay_df %>% dplyr::filter(delay_type == 'Publication'),
  min_year = 2000,
  max_delay = 105,
  xlab = xlab_publish,
  ylab = ylab_publish
)

file.path('viz', 'publication-by-article.png') %>%
  ggplot2::ggsave(plot = gg, width = 5.5, height = 0.75 * 5.5)
file.path('viz', 'publication-by-article.pdf') %>%
  ggplot2::ggsave(plot = gg, width = 5.5, height = 0.75 * 5.5)

plot_journal_history <- function(delay_df, journal_abbrev, xlab, ylab) {
  # Plot all delays for a single journal and delay_type
  x_box = boxplot.stats(delay_df$date_decimal, coef = 1.5)
  xlimits = x_box$stats[c(1, 5)]
  if(length(x_box$out) >= 35) {
    xlimits = boxplot.stats(delay_df$date_decimal, coef = 2.75)$stats[c(1, 5)]
  }
  ylimits = c(0, boxplot.stats(delay_df$delay, coef = 2.8)$stats[5])
  alpha = log(nrow(delay_df)) ^ -1.03 - 0.02
  gg = delay_df %>%
    ggplot2::ggplot(aes(x = date_decimal, y = delay)) +
    ggplot2::geom_point(alpha=alpha, color='#e60000', size=0.8) +
    ggplot2::geom_smooth(linetype=0, fill='#454343', alpha=0.26) +
    ggplot2::coord_cartesian(xlim = xlimits, ylim = ylimits) +
    ggplot2::theme_bw() +
    ggplot2::xlab(xlab) +
    ggplot2::ylab(ylab) +
    ggplot2::theme(plot.margin=grid::unit(c(5, 12, 2, 2), 'points')) +
    ggplot2::annotate('text', x = Inf, y = Inf, label = journal_abbrev, fontface='italic', hjust=1.12, vjust=2)
  return(gg)
}

property_map = list(
  Acceptance = list(directory = 'accept', xlab = xlab_accept, ylab = ylab_accept),
  Publication = list(directory = 'publish', xlab = xlab_publish, ylab = ylab_publish)
)

for (i in 1:nrow(journal_subset_df)) {
  # Iterate through journal-delay_type pairs
  delay_type_ = journal_subset_df$delay_type[i]
  props = property_map[[delay_type_]]
  nlm_id = journal_subset_df$journal_nlm_id[i]
  journal_abbrev = journal_subset_df$journal_abbrev[i]
    
  gg = delay_df %>%
    dplyr::filter(delay_type == delay_type_) %>%
    dplyr::filter(journal_nlm_id == nlm_id) %>%
    plot_journal_history(
      journal_abbrev = journal_abbrev,
      xlab = props$xlab,
      ylab = props$ylab
    )
  
  path = file.path('viz', 'journal', props$directory, sprintf('%s.png', nlm_id))
  ggplot2::ggsave(filename = path, plot = gg, width = 5.5, height = 3.8, dpi = 150)
}

get_beta = function(df) {
  # Regress delay against date and return model information
  model = lm(delay ~ date_decimal, data = df)
  model_df = broom::tidy(model)
  row_df = dplyr::data_frame(
    slope = model_df$estimate[2],
    slope_as_percent = 100 * model_df$estimate[2] / mean(df$delay),
    p_value = model_df$p.value[2]
  )
  return(row_df)
}

slope_df = journal_subset_df %>%
  dplyr::group_by(journal_nlm_id, delay_type) %>%
  dplyr::inner_join(delay_df) %>%
  dplyr::do(get_beta(.))

slope_df %>%
  format(digits=3, scientific=3) %>%
  readr::write_tsv(file.path('data', 'slopes.tsv'))

# Violin plots of per-journal change in delays
gg = slope_df %>%
  ggplot2::ggplot(aes(x = delay_type, y = slope)) +
  ggplot2::geom_violin(linetype=0, fill='#80a5f9') +
  ggplot2::stat_summary(fun.y=mean, geom='point', size=6, color="white", shape='+') +
  ggplot2::stat_summary(fun.y=quantile, geom='point', size=6, color="white", shape='|') +
  ggplot2::geom_hline(yintercept = 0, linetype = 'dashed') + 
  ggplot2::theme_bw() +
  ggplot2::ylab('Δ days of delay per year') +
  ggplot2::scale_x_discrete(name = NULL) +
  ggplot2::coord_flip(ylim = c(-20, 20)) +
  ggplot2::theme(plot.margin=grid::unit(c(2, 2, 2, 2), 'points'))

file.path('viz', 'slope-distributions.png') %>%
  ggplot2::ggsave(plot = gg, width = 5.5, height = 2.3)
file.path('viz', 'slope-distributions.pdf') %>%
  ggplot2::ggsave(plot = gg, width = 5.5, height = 2.3, device = cairo_pdf)