Plot for showing the proportion of missing data per form/country

[aghaynes]

Is your feature request related to a problem? Please describe.

Describe alternatives you've considered

inst <- comp %>%
  group_by(institute.name) %>%
  mutate(n_pat = length(unique(.data$record_id))) %>%
  group_by(institute.name, form) %>%
  summarize(n_pat = max(n_pat),
            n_var = max(N),
            n = sum(n),
            N = sum(N)
            ) %>%
  mutate(pct = n/N*100) %>%
  ungroup()

xval <- inst %>%
  group_by(form) %>%
  summarize(width = log(max(n_var))) %>%
  arrange(match(form, names(dfs))) %>%
  mutate(x1 = cumsum(width),
         x0 = x1 - width,
         lab_x = x0 + width/2) %>%
  select(form, x1, x0, width, lab_x)

yval <- inst %>%
  group_by(institute.name) %>%
  summarize(height = log(max(n_pat)+1)) %>%
  arrange(-order(institute.name)) %>%
  mutate(y1 = cumsum(height),
         y0 = y1 - height,
         lab_y = y0 + height/2) %>%
  select(institute.name, y0, y1, height, lab_y)

p <- inst %>%
  full_join(xval, by = "form") %>%
  full_join(yval, by = "institute.name") %>%
  ggplot(aes(xmin = x0, xmax = x1, ymin = y0, ymax = y1, fill = pct)) +
  geom_rect() +
  myggtheme() +
  viridis::scale_fill_viridis(option="viridis"
                              , direction = -1
                              ) +
  # scale_fill_continuous(type = "viridis") +
  # scale_fill_gradientn(colours = viridisLite::mako(256)) +
  labs(fill = "% complete") +
  scale_x_continuous(
    # breaks = xval$lab_x,
    breaks = c(0,xval$x1),
    # labels = xval$form,
    labels = rep("", length(c(0,xval$x1))),
    position = "top") +
  scale_y_continuous(
    breaks = c(0,yval$y1),
    labels = rep("", length(c(0,yval$y1)))
    # breaks = yval$lab_y,
    # labels = yval$institute.name
    ) +
  theme(axis.text.x = element_text(angle = 45, hjust = 0),
        plot.margin = unit(c(4.5,0,0,5), units = "lines")
        )

yrange <- ggplot_build(p)$layout$panel_scales_y[[1]]$range$range
ylabpos <- yrange[2] + (yrange[2] - yrange[1]) * 0.06
xrange <- ggplot_build(p)$layout$panel_scales_x[[1]]$range$range
xlabpos <- xrange[1] - (xrange[2] - xrange[1]) * 0.06
# xlabels
for (i in 1:length(xval$form))  {
  p <- p + annotation_custom(
    grob = textGrob(label = xval$form[i],
                    hjust = 0, vjust = 0,
                    gp = gpar(cex = 0.8),
                    rot = 45),
    ymin = ylabpos,      # Vertical position of the textGrob
    ymax = ylabpos,
    xmin = xval$lab_x[i], # Note: The grobs are positioned outside the plot area
    xmax = xval$lab_x[i])
}
# ylabels
for (i in 1:length(yval$y0))  {
  p <- p + annotation_custom(
    grob = textGrob(label = yval$institute.name[i],
                    hjust = 1, vjust = 0,
                    gp = gpar(cex = 0.8),
                    rot = 0),
    ymin = yval$lab_y[i],      # Vertical position of the textGrob
    ymax = yval$lab_y[i],
    xmin = xlabpos, # Note: The grobs are positioned outside the plot area
    xmax = xlabpos)
}
gt <- ggplot_gtable(ggplot_build(p))
gt$layout$clip[gt$layout$name == "panel"] <- "off"

pdf(paths("fd", "complete_matrix_site.pdf"),
       height = cm2inch(16), width = cm2inch(16))
grid.draw(gt)

dev.off()

[aghaynes]

usefulness? generalizability?

[mbranca]

Like here (I think), this is useful to be done per form (% of completeness per form), like # complete items / # total items. and then average over the site. Correct?

[aghaynes]

probably, site height is proportional to the number of participants from that site and width is proportional to the number of fields on that form