Kenya Census

Data visualization exercise

This notebook uses TidyTuesday 2021/week4 Kenya Census data containing gender, crops and households indicators of the 2019 Kenya Population and Housing Census, from rKenyaCensus courtesy of Shelmith Kariuki.

Load libaries

library(tidyverse)
library(janitor)
library(waffle)
library(ggsci)
library(ggpubr)
library(scales)

Import data

gender <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-01-19/gender.csv')
crops <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-01-19/crops.csv')
households <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-01-19/households.csv')

Clean and merge data

# clean names
crops_df = crops %>% clean_names()
households_df = households %>% clean_names()
gender_df = gender %>% clean_names()

# change county names to upper for joining
households_df$county <- sapply(households_df$county, toupper)
gender_df$county <- sapply(gender_df$county, toupper)

# recode factors for joining
gender_df = gender_df %>% mutate_at(vars(county), ~recode_factor(.,"TOTAL" = "KENYA"))
crops_df <- crops_df %>% rename(county = sub_county) 

# remove white space in county
crops_df$county <- gsub('\\s+', '', crops_df$county)
households_df$county <- gsub('\\s+', '', households_df$county)
gender_df$county <- gsub('\\s+', '', gender_df$county)

# rename NAIROBI to NAIROBICITY in crops_df
crops_df = crops_df %>% mutate_at(vars(county), ~recode_factor(.,"NAIROBI" = "NAIROBICITY"))

# merge three dataframe
data = list(crops_df, households_df, gender_df) %>% 
  reduce(full_join, by = "county")
dim(data)

[1] 48 18

Visualise missing data

• reference: [@joshyam_](https://twitter.com/joshyam_/status/1351281661250396160/photo/1)

# plot
data %>% summarise(across(everything(), ~mean(!is.na(.)))) %>% 
  gather() %>%
  mutate(key= fct_reorder(key, value)) %>%
  ggplot(aes(key, value)) +
  geom_point(size=3, color="#006d77") +
  geom_segment( aes(x=key, xend=key, y=0, yend=value)) + 
  geom_text(aes(label= percent(value)),
            nudge_y=0.07, size=3) + 
  scale_y_continuous(labels= scales::percent) + 
  theme_minimal() +
  theme(
    panel.grid.major.x = element_blank(),
    panel.grid.minor.x = element_blank(),
    panel.grid.major.y = element_blank(),
    legend.position="none"
  ) +
  theme(axis.text.x=element_blank()) +
  labs(x="", y="% of data present") + 
  coord_flip()

Distribution of Households Growing Permanent Crops by Type and County

crops_df2 <- crops_df %>% mutate_if(is.factor, function(x) tolower(as.character(x)))
crops_df2$county = str_to_title(crops_df2$county)

k_crops = crops_df2 %>% filter(county=="Kenya") 
k_crops_key = k_crops %>% gather("farming":"khat_miraa", key="crop_type",value="households") %>% mutate(households_mil = round((households/1000000),2))

ggdotchart(k_crops_key, x = "crop_type", y = "households_mil",
           color = c("#457b9d"), # Custom color palette
           sorting = "descending",                       # Sort value in descending order
           rotate = TRUE,                                # Rotate vertically
           dot.size = 8,                                 # Large dot size
           label = round(k_crops_key$households_mil,1),
           font.label = list(color = "white", size = 9, 
                             vjust = 0.5),
           ggtheme = theme_pubr()                        # ggplot2 theme
           )+
  theme_cleveland() +                                     # Add dashed grids
    labs(y="Number of households (in millions)", caption= "Data from rKenyaCensus", title="Distribution of Households Growing Permanent Crops by Type", subtitle="2019 Kenya Population and Housing Census")

Crops: Coffee vs Tea

• Inspired by: izz_m2

# prepare data
ct_key = crops_df2 %>% select(county, coffee, tea) %>% gather(crop_type, households, 2:3) %>% filter(!is.na(households)) %>% filter(county!="Kenya") 
ct_key = ct_key %>% group_by(county) %>% mutate(households_total = sum(households)) %>% ungroup() %>% arrange(desc(households_total))

# plot
ggplot(ct_key, aes(households, reorder(county,households_total))) +
  geom_line(aes(group = county)) +
  geom_point(aes(color = crop_type)) + 
  theme_minimal() + 
  scale_color_d3() + 
  labs(y="County",
       x="Number of households",
       color="", 
       title= "Coffee vs. Tea ", 
       subtitle="Number of farming households in Kenya by County",
       caption= "Data from rKenyaCensus")

Crops: Cashew Nut vs. Coconut

# prepare data
mc_key = crops_df2 %>% select(county, cashew_nut, coconut) %>% gather(crop_type, households, 2:3) %>% filter(!is.na(households)) %>% filter(county!="Kenya") 
mc_key = mc_key %>% group_by(county) %>% mutate(households_total = sum(households)) %>% ungroup() %>% arrange(desc(households_total))

right_label <- mc_key %>%
        group_by(county) %>%
        arrange(desc(households)) %>%
        slice(1)

left_label <- mc_key %>%
        group_by(county) %>%
        arrange(desc(households)) %>%
        slice(2)

# plot
ggplot(mc_key, aes(households, reorder(county,households_total))) +
  geom_line(aes(group = county)) +
  geom_point(aes(color = crop_type), size=3) + 
  geom_text(data = right_label, aes(color = crop_type, label = round(households, 0)),
                  size = 3, hjust = -.5) +
  geom_text(data = left_label, aes(color = crop_type, label = round(households, 0)),
                  size = 3, hjust = 1.5) +
  theme_minimal() + 
  theme(legend.position = "bottom") +
  scale_color_nejm() + 
  scale_x_continuous(limits = c(-5000, 50000)) + 
  labs(y="County",
       x="Number of households",
       color="", 
       title= "Cashew nut vs. Coconut ", 
       subtitle="Number of farming households in Kenya by County",
       caption= "Data from rKenyaCensus")

Crops: Cashew Nut vs. Coconut

# waffle plot
# cashew nut
cas_df = crops_df2 %>% select(county, cashew_nut) %>% filter(!is.na(cashew_nut)) %>% arrange(desc(cashew_nut))
cas_df = cas_df[-1,] #remove first row

cas_vec = c(`KILIFI (27940)`=27940, `KWALE (22803) `=22803, `LAMU (8085) `=8085,
    `TANA RIVER (1691) `=1691, `MOMBASA (602) `=602, `TAITA/TAVETA (543) `=543)
    
p1 = waffle(cas_vec/500, rows=8, size=0.6,
      colors=c("#e85d04", "#264653", "#e9c46a","#00509d", "#90be6d", "#9e2a2b"),
      title="Households Growing Cashew Nut by Counties of Kenya",
      xlab="1 square = 500 households, data from rKenyaCensus", pad=7) 
p1 = p1 + theme(plot.title = element_text(size = 12))

# coconut
coco_df = crops_df %>% select(county, coconut) %>% filter(!is.na(coconut)) %>% arrange(desc(coconut))
coco_df = coco_df[-1,] #remove first row
coco_vec = c(`KILIFI (47561)`=47561, `KWALE (31954) `=31954, `LAMU (5017) `=5017,
    `TAITA/TAVETA  (2504) `=2504, `TANA RIVER (2228) `=2228, `MOMBASA (1688) `=1688)
    
p2 = waffle(coco_vec/500, rows=8, size=0.6,
            colors=c("#e85d04", "#264653", "#e9c46a","#9e2a2b", "#00509d", "#90be6d"),
            title="Households Growing Coconut by Counties of Kenya",
            xlab="1 square = 500 households, data from rKenyaCensus")
p2 = p2 + theme(plot.title = element_text(size = 12))

# plot
iron(p1,p2)

Avocados

• Inspired by [@adriaaaaaaan](https://twitter.com/adriaaaaaaan/status/1351665628013355008/photo/1)

avo_df = crops_df2 %>% dplyr::select(county, avocado, farming) %>% filter(!is.na(avocado)) %>% mutate(ratio = round((avocado/farming),4))

avo_df %>% ggplot(aes(reorder(county,ratio), ratio)) + geom_point(color="#606c38",size=2.5) + coord_flip() + 
  scale_y_continuous(labels = scales::percent, limits=c(0, 0.45)) +
  theme_minimal() +
  theme(
    panel.grid.minor.x = element_blank()) +
  labs(x="",y="", title="Avocados from Kenya", subtitle="Percentage of county's farming households farming avocados", caption="Data from rKenyaCensus")

Avocados (map)

• reference: [Kenya X Coffee by @pyyxxo](https://twitter.com/pyyxxo/status/1351568045714645000/photo/1)

library(Hmisc)
library(sf)
library(colorspace)
library(ggrepel)
library(viridis)
library(rKenyaCensus)

  k_shp <- rKenyaCensus::KenyaCounties_SHP

# clean
clean_county <- function(X) {
    X %>%
      tolower %>%
      str_replace_all("[^[:alpha:]]+", "") %>%
      str_replace_all(fixed("city"), "")
}

households = households %>% mutate(County = clean_county(County))
crops = crops %>% mutate(County = clean_county(SubCounty))
gender = gender %>% mutate(County = clean_county(County))

k_shp %>%
    as("sf") %>%
    dplyr::select(-Population) %>%
    mutate(County = clean_county(County)) %>%
    inner_join(households, by="County") %>%
    inner_join(gender, by="County") %>%
    inner_join(crops, by="County") ->
    k_data


# centroids
k_data$centroids <- st_transform(k_data) %>% 
    st_centroid() %>% 
    st_transform(., '+proj=longlat +ellps=GRS80 +no_defs') %>%
    st_geometry()

# centriods as dataframe
k_data_2 <- k_data %>% st_centroid() %>%  
    as_Spatial() %>%                
    as.data.frame()

# select four countries to label 
topavo <- subset(k_data_2, Avocado>50000)
topavo$County <- capitalize(topavo$County)
topavo[, c("County", "coords.x1", "coords.x2")]

topavo$County[topavo$County=="Muranga"] <- "Murang'a"

ggplot(k_data) + 
    geom_sf(aes(fill=Avocado), color="black") +
   geom_text_repel(data = topavo, aes(x = coords.x1, y = coords.x2, label = County), 
                   size=8, color="firebrick4", 
                             #  bungoma  / kakamega / kisii / muranga
                   nudge_x = c(-100000,   -250000,  +5000, -50000), 
                   nudge_y = c(+100000, -250000, -200000, -400000) ) +
   labs(title = "KENYA: Population growing/farming avocados by county", 
        subtitle="",
        fill="",
        caption="Data from rKenyaCencus") +
  scale_fill_continuous_sequential(palette = "Greens") +
   theme_void() + 
   theme(legend.position = "left",
         legend.text=element_text(size=16),
         text = element_text(color="black"),
         plot.caption  = element_text(size = 14, hjust=.5), 
         plot.title    = element_text(size = 32, hjust=0), 
         plot.subtitle = element_text(size = 24, hjust=0, color="gray30"), 
         plot.margin=unit(c(0.5, 0, 0.5, 0),"cm") )

Clustering household data by county

• Reference: [@OzancanOzdemir](https://twitter.com/OzancanOzdemir/status/1351794062626578432/photo/1)

# libraries
library(factoextra)
library(ggthemes)

# prepare data 
cdf<-as.data.frame(households) #change class
rownames(cdf)<-households$County
cdf<-cdf[,-1] # drop first col
cdf<-cdf[-1,] # drop national data

# scale
cdf_scaled = scale(cdf)
head(cdf_scaled)

            Population NumberOfHouseholds AverageHouseholdSize
mombasa      0.2719883          0.5119992           -1.2801769
kwale       -0.2126755         -0.3634175            0.8686915
kilifi       0.6366411          0.1709959            0.6424948
tanariver   -1.0063075         -0.8109828            0.4162981
lamu        -1.2583862         -0.9401290           -0.6015869
taitataveta -0.9756191         -0.6907594           -0.8277836

# check optimal clusters using elbow plot
fviz_nbclust(cdf, kmeans, method = "wss")

# K means with 4 clusters
set.seed(123)
k4 = kmeans(cdf, 4,nstart=25)

# plot
fviz_cluster(k4, data = cdf)+theme_fivethirtyeight()+theme(rect = element_rect(fill = "White",linetype = 0, colour = NA))+labs(title="Cluster of Counties in Kenya based on Household",caption="Data from rKenyaCensus")