Visualizing population growth rate-Haivo

Get the data

Data came from Vietnam general statistic organization To be able to easily access the data from R. I build a simple package gsovn to scrape data from website.

library(tidyverse)
data = gsovn::gso_read(gsovn::avail_dataset$link[30])
data$data %>%
  head(10) %>% 
  knitr::kable()

	2005	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	Prel. 2020
WHOLE COUNTRY	1.17	1.09	1.07	1.06	1.07	1.05	1.08	1.07	1.08	1.12	1.11	1.11	1.17	1.15	1.14
Red River Delta	0.90	0.63	1.28	0.74	1.19	1.08	1.04	1.02	1.09	1.37	1.41	1.38	1.47	1.48	1.33
Ha Noi	2.02	1.37	NA	1.41	2.50	1.93	1.54	1.63	1.70	2.03	2.11	1.99	2.23	2.27	1.89
Ha Tay	2.03	1.14	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA	NA
Vinh Phuc	1.03	0.69	NA	0.66	0.72	0.38	1.09	0.68	1.22	1.26	1.45	1.54	1.36	1.45	1.42
Bac Ninh	0.80	0.95	0.87	0.82	1.73	1.84	2.10	2.06	2.08	3.52	3.23	3.17	3.05	3.08	2.94
Quang Ninh	1.33	1.19	1.12	0.97	0.97	0.93	0.83	0.82	1.01	1.13	1.58	1.90	1.46	1.61	0.96
Hai Duong	0.30	0.30	0.36	0.35	0.56	0.78	0.69	0.58	0.65	0.95	1.11	0.97	1.46	1.02	1.05
Hai Phong	0.89	0.98	0.98	0.89	0.94	1.19	1.29	1.11	1.08	0.96	0.80	0.81	0.75	0.83	1.00
Hung Yen	0.60	0.43	0.44	0.21	0.33	0.54	0.50	0.66	0.56	1.12	1.12	1.10	1.10	1.08	1.06

This dataset contain data of population growth (by percent) in Vietnam. The data is quite in a good a shape, but still need some cleaning

Cleaning

First I need to remove white space from provinces/cities name.

df = data$data
df = df %>% rename(region =1) %>% 
  mutate(region = str_squish(region))

Next I split up provinces/cites by region it belong to.

region = c(
  "WHOLE COUNTRY",
  "Northern midlands and mountain areas",
  "Red River Delta",
  "Northern Central area and Central coastal area",
  "Central Highlands",
  "South East",
  "Mekong River Delta" 
)


region_regex = paste0("(",str_c(region, collapse = "|"),")")

df = df %>% 
  mutate(across(-region, as.numeric)) %>% 
  mutate(group = cumsum(str_detect(region, region_regex))) %>% 
  group_by(group) %>% 
  mutate(group = first(region)) %>% 
  ungroup()

Then pivoting data to the long format

df_long = df %>% 
  pivot_longer(cols = c(-region, -group),names_to = 'year', values_to = "pop_growth_rate")
  
head(df_long,5) %>% 
  knitr::kable()

region	group	year	pop_growth_rate
WHOLE COUNTRY	WHOLE COUNTRY	2005	1.17
WHOLE COUNTRY	WHOLE COUNTRY	2007	1.09
WHOLE COUNTRY	WHOLE COUNTRY	2008	1.07
WHOLE COUNTRY	WHOLE COUNTRY	2009	1.06
WHOLE COUNTRY	WHOLE COUNTRY	2010	1.07

Prepare data for visualization

I want the audience know which provinces/cities belong to which region, so, I need the map data of Vietnam split by regions. This is quite common task for me, hence I saved the map to my personal package {haitools}. Save data this way allow me to load it even when I am not working on my computer.

library(sf)
library(ggrepel)
theme_set(theme_light())
vn_sf = haitools::small_vnsf

# fix thua thien hue, tp ho chi minh name
df_long = df_long %>% 
  mutate(region = case_when(region == "Thua Thien-Hue" ~ "Thua Thien Hue",
                            region == "Ho Chi Minh city" ~ "Ho Chi Minh",
                            TRUE ~ region))

a = map(vn_sf$geometry, ~ colMeans(chuck(.x,1,1))) 
coord_x = map_dbl(a, 1)
coord_y = map_dbl(a,2)

vn_sf = vn_sf %>% mutate(coord_x = coord_x, coord_y = coord_y)

Vietnam map by region

p1 = df_long %>%  
  distinct(region,group) %>% 
  mutate(group = fct_relevel(group, .env$region[-1] )) %>% 
  inner_join(vn_sf, by = c(region = "province")) %>% 
  ggplot(aes(geometry = geometry, fill = group))+
  geom_sf(color= "white", alpha = .5)+
  guides(fill = guide_legend(ncol = 1))+
  theme_void()+
  labs(fill = "")+
  #geom_text_repel(aes(x = coord_x, y = coord_y, label = region),max.overlaps = 20, size = 3)
  theme(legend.position =c(0.4,0.35), text = element_text(size =16))
p1

Next, I visualizing the distribution of growth rate for each region. There are a interesting package {ggdist} which draw half-eye plot for distribution, compare with the typical box-plot or violin-plot.

df_count_group = df_long %>% 
  filter(!region %in% .env$region) %>% 
  filter(year == 2019) %>% 
  count(group)

df_by_group = df_long %>% 
  filter(!region %in% .env$region) %>% 
  left_join(df_count_group, by = c("group"= "group")) %>% 
  mutate(group = fct_relevel(group, .env$region[-1])) 

df_by_group_n = df_by_group %>% distinct(group, n) %>% 
  mutate(new_name = paste0(group, " (n=", n, ")"))
new_name = df_by_group_n$new_name
names(new_name) = df_by_group_n$group

df_by_group = df_by_group %>% 
  mutate(group = fct_relabel(group, ~new_name[.] ))


df_whole = df_long %>% 
  filter(region == "WHOLE COUNTRY")

df_whole2 = df_whole %>% 
  mutate(group = map(region, ~unique(df_by_group$group))) %>% 
  unnest(group)

p2 = df_by_group %>% 
  ggplot(aes(year, pop_growth_rate, group = group, fill = group))+
  ggdist::stat_halfeye(alpha = .7)+
  geom_line(data = df_whole2,
            aes(year, pop_growth_rate, group = region),
            linetype =2, 
            inherit.aes = FALSE) +
  facet_wrap( ~ group, scales = "free",ncol = 2) +
  guides(fill = "none") +
  scale_y_continuous(labels = scales::percent_format(scale = 1)) +
  theme(panel.grid.major.x = element_blank(),
        panel.grid.minor.x = element_blank(),
        strip.background = element_rect(fill = "#9eba89"),
        strip.text = element_text(colour = "#2d3443", size = 18),
        #plot.background = element_rect(fill = "grey80"),
        #panel.background = (fill ="grey80"),
        text = element_text(size= 18)) +
  labs(
    y = NULL,
    x = NULL
  )

p2

At last, I combine two plot with {patchwork}

library(patchwork)
library(ggtext)
p = p1 + p2 +
  plot_annotation(
    title = "Distribution of population growth rate group by area, from 2005 to 2020",
    subtitle = "**(n)** is number of provinces in each group <br/>
                  *Dash-line* is the average of whole country",
    theme = theme(
      plot.title = element_markdown(size = 22),
      plot.subtitle = element_markdown(size = 18)
    )
  )

To save the images, use ggsave()

knitr::include_graphics("Dist-of-pop-growth.png")

Visualizing population growth rate

Get the data

Cleaning

Prepare data for visualization

CATALOG

FEATURED TAGS

FRIENDS